DNA Synthesis Screening
IGSC Protocol, Sequence of Concern, and Gaps
DNA synthesis screening is the commercial chokepoint where providers check both the ordered sequence and the customer before synthetic DNA ships. Under the IGSC Harmonized Screening Protocol, members screen gene-length orders at a 200 bp floor against sequences of concern and verify legitimacy; residual risk concentrates in shorter oligos, benchtop devices, cell-free expression systems, non-IGSC capacity, and AI-designed sequences that may not match reference databases. The sections below set out the baseline, U.S. policy status after the 2024 OSTP framework and EO 14292, and those five gaps, at the level of governance and evaluation only, not operational evasion methods.
Screening is biosecurity’s most practical intervention point in the synthetic-biology era because information still has to become physical DNA through a provider.
- Understand how DNA synthesis screening works as a biosecurity intervention.
- Evaluate the IGSC Harmonized Screening Protocol (sequence and customer screening).
- Analyze regulatory evolution from 2010 HHS guidance through 2025 Executive Order.
- Identify critical gaps: oligonucleotides, benchtop synthesizers, cell-free expression, capacity outside the consortium.
- Assess customer verification (Know Your Customer) processes.
- Recognize future screening approaches and policy needs.
- What it is: sequence + customer (KYC) screening before commercial gene-length synthesis
- Industry baseline: IGSC Harmonized Screening Protocol v3.0 (≥200 bp today)
- U.S. status (as of Aug 2026): 2024 OSTP Framework under revision after EO 14292; S. 3741 introduced, not enacted
- Five hard gaps: oligo fragmentation, benchtop synthesizers, cell-free expression systems, ~20% capacity outside IGSC, AI-designed sequence evasion
- Not a substitute for: the IGSC PDF, IBBIS tooling, or primary statute/guidance (follow outbound links)
Coverage of biosecurity risks stays at the conceptual level appropriate for education and policy analysis. Consistent with responsible information practices:
- Omitted: Actionable protocols, specific synthesis routes, exact pathogen sequences
- Included: Risk frameworks, governance mechanisms, policy recommendations
For detailed biosafety protocols, consult your Institutional Biosafety Committee and relevant regulatory guidance.
Introduction
DNA synthesis screening is biosecurity’s most practical intervention point in the synthetic biology era. As Synthetic Biology and the Democratization of Biotechnology showed, democratized access to biological design tools means the knowledge to create dangerous pathogens is increasingly available. Knowledge alone is not enough. Physical DNA is still required.
Screening operates at that point. When someone orders synthetic DNA from a commercial provider, the company screens both the sequence (is it dangerous?) and the customer (are they legitimate?). This creates a chokepoint: biological information must pass through synthesis companies to become biological reality.
Orders matching sequences in select agent databases trigger human review, and the customer is asked for institutional biosafety committee approval and a description of intended use before synthesis proceeds. Not every legitimate order passes through without friction, and that friction is the point.
How DNA Synthesis Screening Works
The Two-Layer Approach
DNA synthesis screening uses a defense-in-depth strategy with two complementary layers:
- Sequence Screening: Is the ordered DNA itself dangerous?
- Customer Screening: Is the person ordering it legitimate?
Both must pass for an order to proceed, especially for sequences of concern.
| Source | What it is | What to use it for |
|---|---|---|
| IGSC Harmonized Screening Protocol v3.0 | Industry voluntary baseline (sequence + customer) | Primary industry rulebook |
| IBBIS Common Mechanism (commec) | Free open-source screening software + KYC resources | Tooling / global baseline implementation |
| 2023 HHS Guidance / 2024 OSTP Framework | U.S. provider/user guidance; federal-funding procurement conditions | U.S. policy baseline (OSTP under revision post EO 14292) |
| JHU CHS Gene Synthesis Screening Hub | Compliance tables and provider/manufacturer resources | Operational compliance checklist |
| RAND (e.g. Securing Commercial Nucleic Acid Synthesis) | Policy recommendations, conformity / information-sharing gaps | Decision criteria and legislative options |
| This chapter | Protocol baseline, gaps, and U.S. policy status in one place | Fast orientation map; not a substitute for the rows above |
Sequence Screening: BLAST Against Pathogen Databases
The core technical mechanism is sequence homology search using BLAST (Basic Local Alignment Search Tool). In simplified form, the process runs in six steps:
- Order received: Customer submits DNA sequence for synthesis
- Automated comparison: Provider runs BLAST to compare ordered sequence against Regulated Pathogen Database
- Threshold matching: If similarity exceeds set threshold (typically >80% identity over 200+ bp), triggers flag
- Six-frame translation: Sequence translated into all six reading frames (amino acid sequences) to detect codon-optimized evasion
- Human expert review: Flagged orders reviewed by trained biosecurity personnel
- Decision: Accept, accept with additional customer screening, or reject
The Regulated Pathogen Database (RPD) maintained by IGSC includes:
Customer Screening: Know Your Customer (KYC)
When orders involve sequences of concern, providers verify customer legitimacy through KYC-like processes:
Initial onboarding (most intensive):
- Identity verification (government-backed ID)
- Institutional affiliation confirmation
- Sanctions and watchlist screening
- Publication history check (legitimate researcher?)
- Shipping address validation
For select agent sequences:
- Written description of intended use
- Institutional biosafety committee approval
- Principal investigator verification
- Proof of appropriate containment facilities (BSL-3 or BSL-4 if required)
Periodic re-verification:
- Customer legitimacy rechecked annually or when suspicious patterns emerge
Red flags:
- Evasiveness about identity or affiliation, or about basic details of the stated research
- Suspicious shipping addresses (residential instead of institutional)
- Email domain that does not match the claimed institution
- No publication history alongside orders for complex sequences
- Requests to fragment orders across multiple providers
- Unwillingness to provide intended use documentation
- Payment offers outside standard institutional purchasing
None of these alone proves malicious intent, but each raises a question a legitimate researcher can readily answer.
The IGSC Harmonized Screening Protocol
Origins and Membership
The International Gene Synthesis Consortium (IGSC), formed in 2009, comprises leading commercial DNA synthesis companies. IGSC reports that its members represent approximately 80% of global gene synthesis capacity. The remaining share is not thereby unscreened; what it lacks is a common, auditable standard, a gap discussed below.
IGSC membership is voluntary, which creates both strengths (industry buy-in, rapid adaptation) and weaknesses (no legal enforcement, incomplete coverage).
Protocol Evolution
The IGSC has updated its protocol over time:
- Version 1.0 (2009): Initial framework
- Version 2.0 (2017): Enhanced specificity
- Version 3.0 (September 2024): Current standard with increased uniformity and accuracy
Version 3.0 represents significant improvements in screening specificity and operational guidance.
Sequence Screening Requirements
IGSC members must screen:
Minimum length: 200 base pairs or longer (IGSC v3.0 maintains this threshold; separately, 2023 HHS guidance encourages providers to screen down to 50+ bp, though HHS guidance remains voluntary for non-federally-funded research)
Orders under 200 base pairs are not reliably screened. The current IGSC baseline still focuses on 200+ bp orders, which leaves a real short-fragment gap because oligonucleotides can be assembled into larger constructs. The 2024 OSTP framework tried to narrow this by moving toward 50-nucleotide window screening by October 13, 2026, but that framework is under revision after Executive Order 14292. In practice, readers should treat sub-200 bp screening as an unresolved governance problem, not a solved one.
Process:
- Compare complete sequence against RPD
- Screen against internationally coordinated databases (NCBI/GenBank, EBI/EMBL, DDBJ)
- Translate all six reading frames into amino acids
- Screen protein sequences against RPD-derived proteins
- Flag matches based on identity thresholds
Oligonucleotide pools: Protocol recommends biosecurity screening for pooled oligonucleotides (though implementation varies by provider)
Automated initial screening → Human expert review → Decision
Legitimate researchers reasonably ask what screening costs them. Routine orders that produce no sequence match clear automated screening as part of normal order processing. Orders that flag go to human expert review and require documentation from the customer, which adds time before synthesis begins. Published figures for typical review turnaround and any associated verification fees are thin, and practice varies by provider, so researchers should ask their supplier rather than assume a number. The asymmetry is the design intent: documentation a credentialed researcher can supply in a few emails is a barrier to someone who cannot supply it at all.
Customer Screening Requirements
For all orders:
- Customer identification data (name, shipping address, institution, country, phone, email)
- Verification of legitimacy during onboarding
For select agent/toxin sequences:
- Suppliers only sell to legitimate government, university, non-profit, or industrial labs
- Significant portions of regulated genomes require written intended use description
- Institutional verification
Record Keeping
IGSC members retain for minimum 8 years:
- Every synthesized gene sequence
- Vector information (if applicable)
- Recipient identity and shipping address
- Every gene sequence screening result
This creates an audit trail for post-incident attribution and helps identify patterns of concerning orders.
Order Refusal and Reporting
Members reserve the right to:
- Refuse any order without explanation
- Notify other IGSC members about problematic orders
- Report to law enforcement or intelligence authorities
Some companies maintain relationships with FBI Weapons of Mass Destruction Directorate for consultation on ambiguous cases. Certain orders may also trigger Bureau of Industry and Security (BIS) export control reviews if they involve controlled dual-use items under the Export Administration Regulations.
Regulatory Context
2010 HHS Guidance: Foundation and Limitations
In 2010, the U.S. Department of Health and Human Services issued its Screening Framework Guidance for Providers of Synthetic Double-Stranded DNA. This voluntary framework established:
Sequence screening:
- Focus on synthetic double-stranded DNA (dsDNA)
- 200 base pair minimum screening threshold
- Comparison against select agent lists
Customer screening:
- Identity verification
- Institutional legitimacy checks
- Compliance with export control regulations
Purpose: Prevent malicious actors from obtaining genetic material to create select agents by bypassing existing regulations.
Critical limitation: The guidance was voluntary. No legal requirement, no enforcement mechanism, only industry goodwill.
For screening policy analysis, RAND’s 2024 report on securing commercial nucleic acid synthesis reviews government guidelines, industry consortium requirements, and legislative proposals. The report provides recommendations for implementing screening policies, including decision criteria and conformity assessment tools for providers.
The 2010 framework predates:
- Benchtop DNA synthesizers (2015+) enabling on-site synthesis without ordering from providers
- Widespread CRISPR genome editing (2012+) allowing assembly of fragments into functional genomes
- AI protein/genome design tools (2020+) creating novel sequences without homology to known threats
- Expanded RNA synthesis capacity for mRNA therapeutics (2020+)
The guidance focuses on double-stranded DNA from centralized providers. Single-stranded DNA, RNA synthesis, oligonucleotide fragments (<200 bp), and benchtop devices fall into regulatory gray zones. All four developments postdate the guidance, so its scope no longer matches current synthesis capability.
2023-2024 Framework Updates
HHS released an updated Framework for Nucleic Acid Synthesis Screening in October 2023, followed by the April 2024 OSTP Framework:
Key improvements (2023–2024):
- Expands beyond select agents to broader “Sequences of Concern”
- Encourages screening at 50+ bp (down from 200 bp)
- Addresses oligonucleotide pools and benchtop synthesizers
- Mandatory for federally funded research under the 2024 framework (for purchases on or after April 26, 2025) (OSTP, 2024)
- Federal funding recipients must procure from compliant providers
Remaining gaps:
- Still voluntary for non-federally-funded research
- Limited enforcement mechanisms for private sector
- International harmonization incomplete
May 2025 Executive Order 14292
Executive Order 14292 (“Improving the Safety and Security of Biological Research”) directed federal departments and agencies to revise or replace the 2024 OSTP Framework for Nucleic Acid Synthesis Screening.
As of August 3, 2026, ASPR S3 states that federal departments and agencies will revise or replace the 2024 framework, and the page will be updated once a new framework is available (ASPR S3).
In January 2026, the Senate introduced the bipartisan Biosecurity Modernization and Innovation Act of 2026 (S. 3741) to improve nucleic acid synthesis security (see Toward Mandatory Screening). The bill was read twice and referred to the Senate Committee on Commerce, Science, and Transportation (GovInfo).
Critical Gaps in Current Screening
Gap 1: Oligonucleotide Fragments
Short oligonucleotides (typically under 200 base pairs) are often unscreened or minimally screened. The problem: they can be assembled into longer dangerous sequences using standard molecular biology techniques.
The “fragmentation attack” scenario:
- Malicious actor designs dangerous sequence (e.g., virulence factor, toxin gene)
- Fragments sequence into overlapping oligonucleotides, each <200 bp
- Orders fragments from different suppliers (no single order looks dangerous)
- Assembles fragments in their own lab using Gibson assembly, Golden Gate, or similar methods
- Result: functional dangerous gene without triggering screening
NTI and NSABB (National Science Advisory Board for Biosecurity) analyses have highlighted this vulnerability. A 2025 RAND report on protecting biological materials further identifies critical chokepoints in the biological supply network where monitoring could prevent misuse, noting that fragmented oversight and the rise of DIY biology movements create exploitable gaps requiring coordinated intervention across providers and regulators.
This is not theoretical. A red-team experiment demonstrated that 36 out of 38 commercial providers, including 12 of 13 IGSC members, shipped DNA fragments sufficient to reconstruct the 1918 influenza genome when researchers ordered from a fake organization with no laboratory (Edison et al., 2026). Only one company detected the hazard and requested biosafety documentation. The authors argue that DNA fragments capable of generating select agent pathogens should themselves be regulated as select agents.
A 19 August 2026 Nature Biotechnology method, MOSAIC, assembles overlapping microchip oligonucleotides into more than 1,000 distinct gene fragments in one pot by hybridization rather than enzymatic PCR assembly, with near-zero reported misalignment (Wu et al., 2026). That is assembly throughput on the oligo path Gap 1 already flags, not a screening evaluation and not a pathogen construct. Screening of those oligo pools remains the missing control.
Challenge: Screening every oligonucleotide order is computationally intensive and creates false positives. Legitimate research uses oligonucleotides constantly (primers, probes, CRISPR guide RNAs). Flagging all of them would paralyze molecular biology.
Potential solutions:
- Lower screening threshold to 50 bp (computationally expensive)
- Pattern detection across orders (privacy and data sharing challenges)
- Behavioral monitoring (red flags like many small orders from same customer)
Gap 2: Benchtop DNA Synthesizers, The Fundamental Threat to Chokepoint Strategy
Benchtop DNA synthesizers represent the most serious long-term threat to DNA synthesis screening as a biosecurity strategy. Unlike other gaps discussed in this chapter, benchtop synthesis does not just create a loophole in the current system. It fundamentally undermines the entire chokepoint model that makes centralized screening possible.
Technology Context
Centralized DNA synthesis screening assumes orders flow through commercial providers who submit sequences to IGSC protocols or comply with HHS/OSTP requirements. This model faces a fundamental challenge from emerging benchtop DNA synthesizers that enable on-site, in-house synthesis without external oversight.
Current benchtop synthesis platforms:
DNA Script: Enzymatic DNA synthesis platform enabling on-demand oligonucleotide production in individual laboratories. The SYNTAX system uses template-free enzymatic synthesis, so oligonucleotides need not be ordered from a commercial provider.
Other device developers: A 2024 survey of the category identifies Evonetix and Switchback Systems as pursuing gene-length benchtop synthesis, and notes that Telesis Bio’s BioXP performs assembly of supplied fragments rather than de novo synthesis (Langenkamp, Institute for Progress, December 2024).
Two adjacent product categories are frequently misfiled as benchtop synthesizers and should not be. Ansa Biotechnologies and Twist Bioscience are centralized synthesis providers rather than device vendors: they synthesize and ship DNA, including clonal constructs up to 50 kb in Ansa’s case, so their orders still pass through provider sequence and customer screening. Nuclera’s eProtein Discovery is a benchtop cell-free protein-production system, not a DNA synthesizer. Both categories matter for decentralized biological production; neither bypasses nucleic acid synthesis screening.
These technologies are advancing in length, fidelity, and throughput. The biosecurity-relevant change is narrower than “synthesis is decentralizing”: a growing subset of synthesis can occur without any order reaching a screening provider.
The Screening Bypass Problem
Why benchtop synthesizers break the chokepoint model:
No commercial provider involvement: A determined actor could acquire a benchtop synthesizer, synthesize concerning sequences without external oversight, and bypass the entire IGSC/OSTP screening infrastructure.
One-time customer verification: Unlike commercial synthesis where each order triggers screening, benchtop devices are screened (if at all) only once at the point of purchase. Subsequent sequences synthesized on that device receive no external review.
No sequence audit trail: Commercial providers maintain 8-year records of synthesized sequences for attribution and pattern detection. Benchtop synthesis creates no such centralized record unless manufacturers build it into the device.
Fragmentation enables evasion: Even if device purchase screening exists, an actor could fragment dangerous sequences across multiple legitimate-seeming oligonucleotides, synthesize them on a benchtop device, and assemble them using standard molecular biology techniques (Gibson assembly, Golden Gate cloning).
The attack scenario biosecurity experts worry about:
- Actor establishes legitimate research credentials (or works at legitimate institution)
- Purchases benchtop synthesizer for ostensibly legitimate research purposes
- Device is delivered with minimal or no sequence-level screening capability
- Actor synthesizes oligonucleotide fragments encoding dangerous sequences
- No external provider screens the sequences; no pattern detection across orders
- Actor assembles fragments into functional dangerous construct
- Entire process occurs without triggering any biosecurity checkpoint
Policy and Governance Gaps
Current regulatory landscape for benchtop devices:
The 2010 HHS guidance predates widespread benchtop synthesis technology. The 2023 HHS Framework update acknowledges the challenge but provides limited enforceable requirements. Neither framework currently mandates:
- Embedded screening capabilities in benchtop devices
- Cloud-based sequence screening before synthesis occurs
- Manufacturer-maintained audit logs of synthesized sequences
- Device-level controls preventing synthesis of flagged sequences
This creates a fundamental gap: the chokepoint strategy requires centralized screening, but technology is rapidly decentralizing synthesis capability.
Potential Solutions: The IBBIS Common Mechanism
IBBIS (International Biosecurity and Biosafety Initiative for Science, launched 2024) now operates the Common Mechanism, providing free, open-source screening tools:
Now available:
- commec: Free, distributed, open-source screening software
- Designed as global baseline for nucleic acid synthesis screening
- Supports compliance with national and international requirements
- Includes resilience to AI-generated sequences and subversion attempts
Technical approach:
- Sequence screening against curated databases
- Customer screening resources and guidance
- Can integrate with benchtop devices (manufacturer adoption required)
- Regular maintenance updates for evolving threats
Remaining challenges:
- Benchtop manufacturers must voluntarily adopt (no current legal requirement)
- Cybersecurity risks of cloud-connected lab devices
- Potential for modified or counterfeit devices without screening
- International coordination required (cannot just regulate U.S.-sold devices)
- Enforcement mechanisms for non-compliant manufacturers
Current screening works because dangerous sequences must pass through commercial providers who check them. Benchtop devices move synthesis in-house, eliminating that chokepoint.
How much of total synthesis capacity will shift to benchtop devices is not established, and circulating projections should be read as estimates rather than findings. The governance point does not depend on the share: synthesis that happens without an order reaching a provider is synthesis that provider screening cannot see, and that gap widens with adoption unless screening is built into the devices.
The IBBIS Common Mechanism provides free tools, but adoption remains voluntary. Making it mandatory globally will require significant regulatory coordination across manufacturers, governments, and research institutions.
Gap 3: Cell-Free Expression, the Production-Machinery Gap
DNA synthesis screening governs the template, not the machinery that turns a template into a functional product. Cell-free expression (CFE) systems, cell lysates that are freeze-dried for shelf-stable, on-demand protein production, close that step with no customer verification. The dual-use reach is documented in the primary literature: infectious poliovirus, a non-enveloped mammalian virus, was synthesized de novo in a cell-free HeLa extract decades ago (Molla et al., 1991); whole bacteriophages (T7, MS2, and ΦX174) have been assembled from their genomes in one-pot cell-free reactions (Shin, Jardine, and Noireaux, 2012; Rustad et al., 2017); and virus-like particles are a routine cell-free product (Bundy, Franciszkowicz, and Swartz, 2008). Commercial kits now market toxin expression and phage rebooting as features.
A 2026 policy analysis argues these capabilities fall outside both BSL rules and DNA synthesis screening, and proposes capability-based tiering alongside a Cell-Free Expression Oversight Consortium, a know-your-customer body modeled on the IGSC and aligned with export controls (Chitayat and Musunuri, 2026).
Safety-by-design is possible: cell-free systems running a reassigned genetic code can function as a genetic firewall against viral infection and horizontal gene transfer, though this has been demonstrated in research settings rather than deployed (Nyerges et al., 2023; cell-free swapped-code system, 2024).
Gap 4: Capacity Outside the Consortium
IGSC reports that its members represent approximately 80% of global commercial gene synthesis capacity. Nonmembership is not the same as absence of screening: providers outside the consortium may screen under national requirements, customer contract terms, or free tools such as the IBBIS Common Mechanism and SecureDNA. What the remaining share lacks is a common, auditable standard, not necessarily any screening at all.
The measured problem is heterogeneity rather than a clean screened-or-unscreened split. Interviews with 18 synthetic DNA providers and stakeholders found screening length thresholds ranging from 20 to 200 nucleotides across respondents, and few defined procedures for reporting suspicious orders to law enforcement. All three non-IGSC respondents considered IGSC protocols inapplicable to their niche, and two reported following no relevant regulations at all (Kane & Parker, 2024).
Where coverage is weakest:
- Smaller synthesis companies, especially outside the U.S. and Europe
- Academic core facilities offering synthesis services
- Providers in countries without established biosecurity frameworks
- Nontraditional providers who consider existing protocols inapplicable to their work
A customer refused by one provider can approach another operating to a different standard.
Solutions:
- Mandatory screening requirements (requires legislation)
- International harmonization of standards
- Expansion of IGSC membership
- Customer education (reputable researchers use IGSC members)
Gap 5: AI-Designed Novel Sequences
Current screening relies on BLAST-based sequence homology, comparing ordered sequences against databases of known pathogens. A 2025 computational study moved this gap from theoretical to demonstrated for unpatched tools; a 2026 follow-up then tested whether the patches held.
The “Paraphrase Project,” published in Science by Wittmann et al., showed that open-source AI protein design tools (EvoDiff) could redesign known toxins into sequence-divergent variants that unpatched screening missed while models predicted the structures would be preserved. Pre-patch, one screening tool flagged only 23% of AI-generated variants. Working with IGSC and major synthesis providers (Twist Bioscience, IDT) over 10 months, the researchers developed patches now deployed globally. Post-patch detection improved to 72% average, including 97% of sequences most likely to generate functional toxins.
A 2026 follow-up tested fragmented synthetic homologs against four patched tools. Two detected fragments as short as 50 nucleotides, exceeding what the U.S. Framework for Nucleic Acid Synthesis requests, and upgraded versions of the other two improved performance. The authors concluded that the patched tools were effective against the AI-assisted design capabilities tested, while stressing the urgency of developing alternative screening approaches as those capabilities advance (Wittmann et al., 2026).
Key findings:
- No physical proteins were produced in either study; both were computational
- Predicted structural similarity does not establish retained biological function
- The residual 3% of undetected variants is a sensitivity figure, not a demonstration that those sequences would function as toxins
- Detection declines below tool-specific sequence-similarity boundaries, which is the open research question
- Function-aware and contextual screening remain the longer-term need
Emerging approaches to close this gap:
- Function-based screening: Predict protein function from sequence, flag dangerous functions even without homology
- Machine learning classifiers: Train models on dangerous vs. safe sequences, detect novel threats
- Protein structure prediction: Use AlphaFold-like models to infer function from predicted structure
A 2026 perspective argues that sequence-based and function-based screening should be treated as complementary, with phased implementation starting with tractable targets like toxins before expanding to more complex pathogenic functions (Abel et al., 2026).
A 2025 NIST inter-tool analysis adds a useful baseline for the screening layer itself. NIST constructed a blinded 999-fragment dataset and compared six sequence-screening tools, reporting baseline performance above 95% sensitivity and 97% accuracy, with disagreements traced largely to differences in sequence-of-concern definitions or algorithmic methods (Laird et al., 2025). This supports harmonized benchmark datasets as governance infrastructure, but it does not solve the harder problem of screening AI-designed proteins by function rather than sequence similarity.
SecureBio’s September 2026 Astra assessment underscores why nucleic-acid order screening remains load-bearing even when model safeguards look strong. A railfree agentic run reached state-of-the-art on Advanced Screening Evasion against commercial screening algorithms, while the pre-release checkpoint with system-level bio-classifiers disabled refused the task entirely (n=100) (SecureBio, 2026). Capability under elicitation and refusal by a checkpoint under test are different measurements. Physical-order screening still closes the path from digital design to purchasable DNA when model filters fail, are stripped, or are bypassed out of distribution. Only the outcome and the refusal result are reported here; evasion methods are not reproduced.
A complementary provenance approach is SynthIDBio (Stutz et al., Nature, 2026): function-preserving sequence watermarks embedded during ProteinMPNN-style design, with watermarked binders retaining affinity comparable to non-watermarked controls (targets including VEGF-A, SARS-CoV-2 RBD, and PD-L1) and near-perfect detection, plus an optional structure watermark via fine-tuned AlphaFold 3. Resequencing or redesign can erase the sequence mark; open-weight bypass remains possible; this is a proof-of-concept for provenance, not a substitute for synthesis screening. Google DeepMind’s Introducing SynthID Bio announcement (30 Sep 2026) keeps that Nature paper as the methods cite of record while teaching the operational layer: biosecurity as a Swiss-cheese stack where model safeguards, customer vetting, and design-level watermarks cover each other’s gaps. Sarah Carter (Science Policy Consulting) frames the mark as linking designs to the model developer so synthesis providers can streamline screening for customers who used those models; James Diggans (Twist Bioscience) calls watermarking a promising addition that could focus review on sequences that warrant closer look. The same announcement notes optional AlphaFold 3 structure watermarking (fine-tune part of the diffusion network so predicted coordinates carry a detectable signature) and early collaboration with the Hie lab (Stanford / Arc) to watermark DNA of an Evo 2-designed bacteriophage after functional culture tests; treat the phage pipeline as a watch item until a technical manuscript appears. Lab announcement and partner quotes; not a substitute for the Nature methods paper or for nucleic-acid order screening.
A Testable Definition of “Sequence of Concern”
Inter-tool disagreement has a definitional component: organism control lists do not by themselves resolve which individual sequences warrant follow-up. The Sequence Biosecurity Risk Consortium (SBRC) therefore pairs a theoretical definition, the Biosecurity Flag Rubric, with an operational definition, a collection of curated test sets. The rubric returns Flag, No Flag, or Undefined. The test sets separately classify consensus outcomes as Flag, No Flag, Optional Flag, or Disputed. Undefined means the rubric does not yet cover the sequence class; Disputed means participating screening tools disagree (Alexanian et al., 2026).
The version 1.0 collection contained approximately 1.1 million sequences across 42 clusters evaluated by four independently developed screening systems; applying the rubric reduced disputed classifications by 44.3% for controlled viruses and 10.7% across the collection (Alexanian et al., 2026). These figures measure changes in classification agreement within the consortium’s test sets, not screening sensitivity or real-world risk reduction.
The framework also treats false positives as a standards problem. It places selected low-risk gene classes and engineered-construct test sets alongside controlled threats, allowing false-positive behavior to be evaluated rather than treating taxonomic origin as sufficient evidence of risk. Version 1.0 covers human pandemic-potential viruses, selected low-risk gene classes, and controlled toxins; bacterial and fungal coverage, function-based criteria, and jurisdictional differences remain planned extensions. Several authors were employed by participating screening providers, and the underlying sequence collection is access-controlled; these limitations make independent proficiency testing an important next step (Alexanian et al., 2026).
SecureDNA now provides free, privacy-preserving screening for sequences 30+ nucleotides, having screened 67 million nucleotides across providers in the US, Europe, and China (Baum et al., National Science Review, 2026). Their system already exceeds current regulatory requirements and includes protections against AI-designed evasion.
Even PPI prediction tools proposed as inference-time biosafety filters at the AI design layer show similar limitations: AlphaFold 3, AF3Complex, and SpatialPPIv2 all failed to detect experimentally validated SARS-CoV-2 binding mutants. See Inference-Time Filters: A Demonstrated Weakness for details.
Customer Verification in Practice
The KYC Framework for Biosecurity
Know Your Customer (KYC) approaches, adapted from financial services anti-money laundering practices, help verify legitimacy of DNA synthesis customers.
The International Biosecurity and Biosafety Initiative for Science (IBBIS) provides resources for synthesis providers to implement KYC:
Identity signals:
- Government-issued ID verification
- Institutional email domain verification (@university.edu, @institute.org)
- Publication history (ORCID, PubMed, Google Scholar)
- Professional licenses or credentials
Institutional signals:
- Research Organization Registry (ROR) verification
- NIH grant numbers or other funding verification
- Institutional biosafety committee registrations
- Prior order history (established customer?)
Risk scoring:
- BioSecure framework proposes AI-enabled cumulative risk assessment
- Combine multiple verification signals into confidence score
- Higher-risk sequences require higher confidence in customer
A July 2026 evaluation tested five LLMs on information-gathering for legitimacy screening against 41 researcher profiles. The best configuration matched expert-human flag accuracy (90.2% versus 89.0%) at about one-tenth the cost ($1.18 versus $14.04 per customer, including human review) (Acelas et al., 2026). The authors keep ship, follow-up, and reject with a human; this is a lab study on documented researchers, not an in-situ provider trial.
A 2026 policy-and-practice review reports that 2024–2025 OSTP, UK DSIT, IGSC Harmonized Screening Protocol v3.0, and ISO 20688-2:2024 aligned on five customer-screening practices: tiered screening, identity verification for all customers, legitimacy verification for customers ordering sequences of concern, legitimate-use verification for those orders, and supply-chain screening for third-party vendors (Carter, Boldrini, and Alexanian, 2026). IBBIS New Customer and SOC Order forms are one implementation of that consensus. The remaining operational gap is end-user checks when the purchaser is a reseller, academic core facility, or other third-party vendor; the review is a practitioner analysis of tools the authors’ organization publishes, not an independent trial of screening effectiveness.
The “Trust but Verify” Challenge
Customer screening faces a fundamental tension: preventing misuse without creating excessive friction for legitimate research.
Too strict: Legitimate researchers face delays, increased costs, bureaucratic burden. Science slows down.
Too permissive: Malicious actors slip through. Biosecurity fails.
Current approaches err toward permissive for established customers at reputable institutions, strict for new customers or concerning sequences. This makes sense since most research is legitimate, but it creates windows for insider threats or compromised credentials.
Future Approaches and Policy Needs
Toward Mandatory Screening
The April 2024 OSTP Framework makes screening mandatory for federally funded research (effective April 2025), and Executive Order 14292 directs expansion to non-federally-funded settings. But voluntary frameworks have limits:
Arguments for mandatory screening:
- Sets one standard across providers inside and outside the consortium
- Creates level playing field (all providers follow same rules)
- Enables enforcement and penalties for non-compliance
- Sends clear policy signal about biosecurity priority
Arguments for voluntary approaches:
- Industry self-regulation responds faster to technological change
- Mandatory requirements may be overly burdensome
- International harmonization difficult with mandatory national requirements
- Risk of driving synthesis to unregulated jurisdictions
Mandatory screening with international coordination has drawn support from biosecurity policy organizations and, notably, from synthesis providers themselves, as the endorsements of the 2026 Senate bill below show. The challenge is implementation.
In February 2026, Senators Tom Cotton (R-AR) and Amy Klobuchar (D-MN) publicly announced the Biosecurity Modernization and Innovation Act of 2026 (S. 3741). Key provisions:
- Mandatory screening: Directs Secretary of Commerce to require all gene synthesis providers to screen orders and customers
- Dangerous sequence registry: Mandates compilation of a federal list of dangerous sequences for vetting
- NIST governance sandbox: Creates a biotechnology governance sandbox at NIST for testing biosecurity tools and enabling adaptive policymaking
- OSTP gap assessment: Gives OSTP 90 days to identify oversight gaps and develop an implementation plan
The bill has broad support from both policy organizations (NTI, FAS, National Security Commission on Emerging Biotechnology) and synthesis providers themselves (Twist Bioscience, IDT, Ginkgo Bioworks). Industry endorsement of mandatory regulation of their own sector is notable, signaling recognition that voluntary frameworks have reached their limits.
The bill remains in the Senate and has not been enacted.
California has made the federal screening framework binding on the companies themselves. AB 1864 (Berman, Chapter 852) was approved by the Governor on September 30, 2026. It bars a manufacturer from producing benchtop nucleic acid synthesis equipment in California, or selling or delivering it to a customer there, and bars a provider from producing or shipping synthetic nucleic acids subject to screening, unless the company adheres to the September 2024 revision of the Framework for Nucleic Acid Synthesis Screening, attests to that compliance, and posts the attestation publicly. Where the framework says “should,” the statute reads it as a requirement; where it says “encouraged,” it stays a recommendation. The Attorney General may recover a civil penalty of up to $5,000 per day a violation continues, and the chapter does not reach customers who are neither providers nor manufacturers.
Two details matter for anyone tracking the federal picture on this page. The statute names the September 2024 version as the sole operative framework, so a federal revision or replacement under Executive Order 14292 does not flow into California law unless the Legislature amends the definition. And framework provisions that take effect on or after October 13, 2026 have no effect under the state chapter until October 13, 2027. State law now supplies the enforcement that the voluntary federal framework lacks, but only inside one state and only against the 2024 text.
In June 2026, an open letter at screendna.org calling for mandatory U.S. nucleic acid synthesis screening and recordkeeping was co-organized by the Foundation for American Innovation and the Institute for Progress and signed by frontier AI lab leaders (including Google DeepMind, OpenAI, Anthropic, and Microsoft AI), synthesis-industry leaders (including Twist Bioscience and Ansa Biotechnologies, with IGSC board chair Diggans among the signers), life-sciences figures (including David Baker and Tom Inglesby), and former White House biosecurity officials. The letter treats screening of DNA orders and synthesis equipment, plus order and sequence recordkeeping for traceability, as a long-standing biosecurity measure whose urgency rises with AI capability, while stating that support for the mandate does not depend on any particular AI-risk worldview. Advocacy coalition letter and legislative-pressure artifact, not enacted law; pair with this chapter’s S. 3741 and House Science hearing material.
A September 2026 CRS In Focus brief (IF13269) maps the same voluntary-versus-mandatory stack for Congress: IGSC protocols and lab-side frameworks beside open calls for mandatory nucleic-acid order screening and 119th Congress bills that would harden standards and authorities (CRS, 2026). Staff briefing: not new empirical misuse rates.
The same 16 September 2026 House Science hearing put mandatory / verifiable nucleic-acid order screening at the center of witness recommendations (hearing page). Franz argued that no binding federal requirement currently obliges all synthesis providers to screen orders and verify customers, and that Congress should establish that requirement (Franz testimony PDF). Hodges used the natural-gas odorization analogy: the United States did not abandon piped gas because leaks were dangerous; it required a cheap universal signal so a leak announces itself. In that framing, verifiable screening of synthetic-DNA orders is the odorant for scaled biotechnology (Hodges testimony PDF).
Hearing recommendations are not a signed federal screening statute. They strengthen the policy case for a uniform provider standard; they do not themselves create enforceable Commerce Department rules or close benchtop and oligo gaps already mapped on this page.
The same Science issue that published King et al. carried a Johns Hopkins Center for Health Security Perspective stating that generation of functional viral genomes has urgent biosafety and biosecurity implications (Inglesby and Hanke, 2026). As independently reported from that Perspective, the authors wrote that the ability to compose viral genomes using generative AI now exists and that governance to steer it does not, and they said work on pathogens that could infect humans, animals, or plants should not be pursued because such genomes might encode new pathogens that cannot be contained by existing countermeasures (Sample, The Guardian, August 2026). That is a reported editorial recommendation, not enacted law, and it does not establish design of mammalian or human-pathogenic viruses.
Steph Guerra’s September 2026 Bulletin of the Atomic Scientists opinion treats the same Science phage result as a warning that a usable window remains for layered defenses: mandatory gene-synthesis screening, calibrated access to high-risk sequence datasets, and pathogen early warning with attribution and demonstrated rapid response, aligned with the defense-in-depth package in Guerra et al., 2026, RAND RR-A4999-1 (Guerra, Bulletin, 4 Sep 2026). That is opinion commentary, not new experimental evidence of human-virus design.
Complementing Point-of-Sale Screening with Life-Cycle Oversight
Provider screening governs a transaction: it evaluates an order and the customer before synthesis or sale. It does not, by itself, govern possession, use, or transfer after delivery, or equivalent material produced within an institution. A 2026 RAND report therefore proposes using the Federal Select Agent Program (FSAP) as a complementary life-cycle layer for certain nucleic acids, while retaining provider-focused screening as the point-of-sale control (Sharkey et al., 2026).
The report also identifies implementation requirements. Any expansion should define regulated scope precisely and in machine-readable form, scale obligations to risk, minimize paperwork, protect privacy and intellectual property, and use phased implementation with stakeholder feedback and explicit opportunities to stop or revise later phases. The authors also call for quantitative cost-benefit analysis during implementation rather than assuming that a wider regulatory perimeter produces a net benefit (Sharkey et al., 2026).
RAND’s framework is a policy recommendation, not a current FSAP rule. Its burden and risk-reduction analysis is based on qualitative interviews and technical-policy assessment, without definitive economic modeling or wet-laboratory validation. The report’s legal conclusion about existing statutory authority also requires independent review of primary law before it is treated as settled.
International Harmonization
DNA synthesis is a global industry. A U.S.-only mandatory framework just shifts orders to providers in other countries. Effective screening requires international agreement.
Efforts toward harmonization:
- IGSC: Multinational membership, harmonized protocol
- Australia Group: 42 participating countries plus the European Commission, coordinating export controls
- Biological Weapons Convention: 189 state parties (but no verification mechanism for DNA synthesis)
- IBBIS Common Mechanism: Free, open-source screening tools designed as global baseline
- ISO 20688-2:2024: International standard for production and quality control of synthesized gene fragments includes biosecurity provisions
Progress is slow. Biosecurity competes with commercial interests, national sovereignty concerns, and lack of urgency (no major DNA synthesis-enabled attack has occurred yet).
A 28 July 2026 open-access mini-review treats provider sequence screening and customer verification as necessary but no longer sufficient as standalone safeguards, and organizes residual risk around four interacting dimensions: design capability, access, governance strength, and liability and attribution clarity (Kato et al., 2026). The authors map those profiles onto a four-tier escalation model from maintained baseline controls to internationally supported minimums, and they state that no binding global standard currently requires all providers to apply the same sequence-and-customer procedures, so the oversight an order receives can depend more on provider and jurisdiction than on the order’s risk profile. The four-tier model is a diagnostic, not a scored ranking, and the authors pair a shared baseline with capacity-building rather than identical rules everywhere.
Function-Based Screening
The October 2025 Science paper showed that homology-based screening is sensitive to sequence divergence and needs continual patching to keep pace with generative design tools. The longer-term response is function-based screening: does a sequence do dangerous things, regardless of similarity to anything already catalogued?
Technical approaches:
- Protein function prediction: Use AlphaFold, ESMFold, or similar models to predict structure/function from sequence
- Danger classifiers: Train ML models on dangerous vs. benign sequences, flag novel threats
- Pathway analysis: Identify if sequence encodes steps in biosynthesis of toxins or virulence factors
SecureDNA already incorporates some function-based protections. Broader adoption remains computationally challenging and prone to false positives.
Function-based screening promises detection of dangerous sequences regardless of homology to known threats. Predicting protein function from sequence alone remains difficult, and minor prediction errors create false positives: benign sequences flagged as dangerous. A provider that flags too many legitimate orders loses customers to competitors; a provider that misses real threats fails at biosecurity. Setting the threshold is therefore as much a values judgment as a technical one, trading research friction against security coverage.
How does DNA synthesis screening work and what are the two layers of protection?
DNA synthesis screening uses a defense-in-depth strategy with two complementary layers. Sequence screening compares ordered DNA against Regulated Pathogen Databases (U.S. Select Agents, Australia Group, EU dual-use lists) using BLAST similarity matching. The system translates sequences into all six reading frames to detect codon-optimized evasion attempts. Customer screening verifies identity, institutional affiliation, publication history, and intended use through Know Your Customer processes. For select agent sequences, both layers must pass: the sequence must be justified scientifically and the customer must provide institutional biosafety committee approvals and proof of appropriate containment facilities.
What is the current regulatory status of DNA synthesis screening in the United States?
The regulatory landscape has evolved significantly. The 2010 HHS guidance established voluntary screening for double-stranded DNA over 200 base pairs. The OSTP Framework for Nucleic Acid Synthesis Screening (released April 29, 2024) made screening a condition of receiving federal life sciences research funding for procurements on or after April 26, 2025 (OSTP, 2024). Executive Order 14292 (May 5, 2025) directed revision or replacement of the framework (GovInfo). As of August 3, 2026, ASPR S3 states that federal departments and agencies will revise or replace the 2024 framework, and the page will be updated once a new framework is available (ASPR S3). In January 2026, the Senate introduced the Biosecurity Modernization and Innovation Act of 2026 (S. 3741), which would require the Secretary of Commerce to promulgate regulations to improve nucleic acid synthesis security (GovInfo).
What are benchtop DNA synthesizers and why do they threaten the screening chokepoint model?
Benchtop DNA synthesizers, such as DNA Script’s SYNTAX system, enable on-site DNA synthesis in individual laboratories without ordering from commercial providers. Two adjacent categories are often misfiled here: Ansa Biotechnologies is a centralized synthesis service whose orders still pass through provider screening, and Nuclera’s eProtein Discovery is a benchtop protein-production platform rather than a DNA synthesizer. True benchtop synthesis undermines the chokepoint strategy: instead of each sequence triggering screening at a commercial provider, benchtop devices are screened only once at purchase. Subsequent sequences synthesized on the device receive no external review unless manufacturers build screening into the device. A determined actor could acquire a benchtop synthesizer through legitimate channels, then synthesize concerning sequences without external oversight. The IBBIS Common Mechanism provides free screening software that can integrate with benchtop devices, but manufacturer adoption remains voluntary.
How do AI-designed sequences evade current screening and what solutions exist?
Current screening relies on BLAST-based sequence homology, comparing ordered sequences against databases of known pathogens. In October 2025, Microsoft researchers showed in Science that AI protein design tools could redesign known toxins into sequence-divergent variants that unpatched screening missed, while models predicted the structures would be preserved. No physical proteins were made, so retained biological function was predicted rather than shown. Working with IGSC and synthesis providers, they developed patches now deployed globally that flagged 97% of the sequences most likely to yield functional toxins. A 2026 follow-up found those patched tools effective against the AI-assisted design capabilities tested, with two of four tools detecting fragments as short as 50 nucleotides, while urging development of alternative approaches as detection declines at lower sequence similarity (Wittmann et al., 2026). Complementary solutions include function-based screening that predicts protein function regardless of homology, machine learning classifiers trained on dangerous versus safe sequences, and protein structure prediction using AlphaFold-like models. SecureDNA provides free screening for sequences 30+ nucleotides with protections against AI-designed evasion.
What is the oligonucleotide fragmentation attack and why is it hard to prevent?
The fragmentation attack exploits the 200 base pair screening threshold: a malicious actor could design a dangerous sequence, fragment it into overlapping oligonucleotides each under 200 bp, order fragments from different suppliers, and assemble them using standard molecular biology techniques like Gibson assembly or Golden Gate cloning. No single order looks dangerous, yet the assembled product is functional. RAND, NTI, and NSABB analyses have all highlighted this vulnerability. Prevention is challenging because screening every oligonucleotide order creates computational burden and false positives that would paralyze legitimate research, which constantly requires oligonucleotide orders for primers, probes, and CRISPR guide RNAs. Potential solutions include lowering screening threshold to 50 bp, pattern detection across orders, and behavioral monitoring for red flags like many small orders from the same customer.
Why is international harmonization of screening standards important?
DNA synthesis is a global industry. A U.S.-only mandatory framework would shift orders to providers in other countries without screening requirements, undermining the chokepoint strategy. Effective screening requires international agreement so malicious actors cannot simply shop for unscreened providers in permissive jurisdictions. Current harmonization efforts include the IGSC’s multinational membership with Harmonized Screening Protocol, the Australia Group coordinating export controls across 42 participating countries plus the European Commission, the IBBIS Common Mechanism providing free global baseline screening tools, and ISO 20688-2:2024 international standard including biosecurity provisions. Progress is slow because biosecurity competes with commercial interests, national sovereignty concerns, and lack of urgency since no major DNA synthesis-enabled attack has occurred yet.
What is the Biosecurity Modernization and Innovation Act of 2026?
Bipartisan legislation introduced in the Senate on January 29, 2026 (S. 3741) to require the Secretary of Commerce to promulgate regulations to improve nucleic acid synthesis security. The bill was read twice and referred to the Senate Committee on Commerce, Science, and Transportation (GovInfo).
This chapter is part of The Biosecurity Handbook. For handbook-wide priorities, see the Executive Summary.