Synthetic Biology
Biology has become an engineering medium. The capability is real. The oversight has not kept up.
(down from $3 billion for the original Human Genome Project completed in 2003)
(one approved: Casgevy for sickle cell disease, late 2023)
(still has no verification mechanism, unlike the chemical and nuclear regimes)
A note on framing. This page is about biology as an engineering capability - what can now be built, edited, designed, and synthesised - rather than about pandemic response, which has its own page on this site. The two stories overlap at the dual-use frontier, but they are different: one is about what nature can do to us and how we respond; this one is about what people can now do with biology, who can do it, and what the governance picture looks like. The capability is genuine and is mostly a story about extraordinary scientific progress. The dual-use angles are real and are mostly a story about institutional gaps that have not kept up.
What synthetic biology actually is now
"Synthetic biology" is a loose label that covers several distinct capabilities developed mostly since 2000, all of which share a common idea: biology as something engineerable, with parts that can be designed, modified, and combined in deliberate ways. The term is used by the field itself but the underlying capabilities matter more than the label.
Reading DNA cheaply. Sequencing the first human genome cost about $3 billion across thirteen years and finished in 2003. Today the same sequencing costs about $200 and can be done in a day. The price drop has been faster than Moore's Law and has been the foundation under almost everything else in the field. Population-scale genomic studies, real-time pathogen surveillance, personalised cancer treatment, and consumer ancestry testing are all downstream of cheap reads.
Editing DNA precisely. CRISPR-Cas9, described as a tool for genome editing in a 2012 paper by Jennifer Doudna and Emmanuelle Charpentier (Nobel 2020), made it possible to cut DNA at a specific sequence with reasonable accuracy and at low cost. Successor systems (base editors, prime editors, CRISPR-Cas12, others) have refined what is possible. The first CRISPR-based therapy, Casgevy for sickle cell disease and beta-thalassemia, was approved by the FDA and UK regulators in late 2023. Roughly fifty more CRISPR-based therapies are in clinical trials across cancer, blood disorders, hereditary blindness, and other targets.
Writing DNA from scratch. Gene synthesis - building a stretch of DNA from chemical components without copying it from an organism - has fallen in cost from about $3 per base pair in 2010 to under ten cents per base pair today. Companies like Twist Bioscience, IDT, and GenScript will synthesise arbitrary DNA sequences and ship them within days. Whole bacterial genomes have been synthesised from scratch (the J. Craig Venter Institute synthesised a fully synthetic Mycoplasma in 2010 and a minimal bacterial genome in 2016). The bottleneck on what can be made is shifting from synthesis cost to design.
Designing new proteins. The most-recent capability shift is the ability to design novel proteins that do not exist in nature, computationally. AlphaFold (Google DeepMind, 2020-2024) made it possible to predict the three-dimensional structure of an arbitrary protein from its amino-acid sequence with accuracy comparable to experimental methods. RoseTTAFold and the broader work of David Baker's lab at the University of Washington (Baker shared the 2024 Nobel in Chemistry for protein design) extended this to designing entirely new proteins for chosen functions. Proteins designed in silico and tested in lab now include novel enzymes, vaccine antigens, antibodies, and binders that nature did not produce. The barrier to "I want a protein that does X" has dropped dramatically.
Manufacturing with biology. Engineered microbes are now used commercially to make insulin, growth hormones, antibiotics, vaccines, vitamins, fragrances, dyes, biofuels, and a growing list of other products. Lab-grown meat (cultured from animal cells without slaughter) reached the market in Singapore in 2020 and the US in 2023, though commercial scale remains small. Microbially-produced spider silk, leather alternatives, and structural proteins are at various stages of commercialisation. The list will keep growing.
The dual-use frontier
The most consequential property of synthetic biology is that the same capabilities that enable medicine also enable harm. This is unlike most technology categories, where productive and destructive uses can usually be separated. In biology, the gap is much narrower. The result is a set of policy and oversight problems that the field has been grappling with for two decades and has not solved.
Gain-of-function research. Some pathogen research deliberately enhances transmissibility, virulence, or host range - to study what mutations enable pandemic-level adaptation. The scientific case is that understanding the path makes it easier to anticipate and counter. The biosafety case is that the research itself creates the risk it is studying. The 2014 US moratorium on gain-of-function research on potential pandemic pathogens was lifted in 2017; subsequent revisions to the Dual Use Research of Concern (DURC) framework have produced a regulatory architecture that critics argue is more performative than enforceable. The COVID-19 origin debate sharpened the political stakes substantially. As of 2026 the policy framework remains contested, with named researchers across the spectrum from "more research, more transparency" to "much stricter restrictions, possibly bans" arguing seriously.
Gene-synthesis screening. Commercial gene-synthesis providers can technically receive an order for any DNA sequence. Industry voluntary standards (the International Gene Synthesis Consortium, founded 2009) screen orders against a database of pathogen and toxin sequences, but coverage is partial and bypass is possible. The October 2023 US executive order required federally-funded research to use providers that screen orders, and the 2024 follow-up policy strengthened the regime; the international picture remains fragmented. As gene synthesis costs drop and providers proliferate, the screening problem gets harder rather than easier.
AI-designed pathogens or toxins. A 2022 paper by Urbina, Lentzos, Invernizzi, and Ekins in Nature Machine Intelligence showed that running a standard pharmaceutical AI tool in reverse - asking it to design molecules with maximum toxicity instead of maximum therapeutic value - generated tens of thousands of novel toxic molecules in hours, including known chemical warfare agents and previously-unknown candidates. The paper was published deliberately to draw attention to the dual-use nature of AI drug-discovery tools. The same logic applies to protein design and pathogen engineering. The community-and-policy response so far has been concerned but not coordinated.
Lower technical barriers. The skills needed to do significant biological-engineering work have fallen substantially. A graduate student with access to a standard molecular biology lab can now do experiments that would have required a major facility twenty years ago. The international Genetically Engineered Machine (iGEM) competition, founded at MIT in 2003, has trained thousands of undergraduate teams in synthetic biology. DIY-bio communities operate hundreds of community labs globally. Most of this capability is used for productive purposes; the same training and tools enable harm if misused.
The verification problem. The Biological Weapons Convention was signed in 1972 and has 184 state parties. It bans development, production, and stockpiling of biological weapons. It has no formal verification mechanism, unlike the chemical-weapons regime (OPCW inspections) or the nuclear regime (IAEA safeguards). A protocol with verification provisions was negotiated through the 1990s and rejected in 2001 under the George W. Bush administration; later attempts have not succeeded. The result is that the international architecture against bioweapons depends almost entirely on state self-restraint and intelligence-based attribution after the fact, which is a thin foundation given the technology trajectory.
The therapy frontier (the productive side)
The dual-use story is real, but it would be misleading to lead with it. The dominant story of synthetic biology so far has been productive: clinical and commercial uses arriving at substantial scale.
Gene therapies that work. Casgevy (Vertex/CRISPR Therapeutics) cures sickle cell disease in most treated patients - the first regulatory approval of any CRISPR-based therapy, in late 2023. Luxturna treats a form of inherited blindness. Zolgensma cures spinal muscular atrophy in infants with a single dose. Hemgenix and Roctavian treat hemophilia. The list is longer in 2026 than it was in 2023, and the pace is accelerating. Many of these therapies are extraordinarily expensive (some over $2 million per treatment), which is its own structural problem - but the medical reality is that diseases previously written off as untreatable now have effective treatments.
Cancer immunotherapy. CAR-T cell therapy, where a patient's immune cells are genetically modified to recognise their cancer, has produced durable remissions in some leukaemias and lymphomas. Newer "off-the-shelf" CAR-T approaches and CRISPR-edited variants are extending the technology to more cancer types. The work that won the 2018 Nobel for James Allison and Tasuku Honjo on checkpoint inhibitors started this category; the gene-editing layer on top has pushed it further.
mRNA platform technology. The COVID-19 mRNA vaccines proved a delivery platform that turns out to be more general. mRNA vaccines for cancer (currently in trials), CMV, RSV, and several other targets are advancing. The "100-day vaccine" target promoted by CEPI and others depends on mRNA infrastructure being kept warm rather than rebuilt each time. Whether the political will to maintain that infrastructure persists is a separate question.
Designed enzymes, antibodies, and binders. The Baker lab and a growing number of other groups are now designing proteins for industrial use - enzymes that break down PFAS forever-chemicals, antibodies tuned for specific viral variants, and binders for previously undruggable targets. The implications for industrial chemistry, environmental cleanup, and previously-impossible therapies are substantial. Most of the press attention has gone to AI; this is one of the places where AI-and-biology together are producing concrete capability that did not exist five years ago.
The governance gap
The scientific community has been more aware of dual-use risks than the broader public conversation suggests, but the institutional response has been slower than the capability curve.
What exists. National-level dual-use research oversight in some countries (the US Dual Use Research of Concern policy, the UK Health Research Authority, EU member-state systems with varying rigour). The voluntary gene-synthesis screening standards. Institutional Biosafety Committees in research universities. Export controls on specific pathogens through the Australia Group (a 43-member informal export-control regime). The Biological Weapons Convention, without verification. NTI's Common Mechanism for assessing biosecurity-relevant research. The WHO Pathogens Prioritization framework.
What is missing. Binding international verification of state-level bioweapons compliance. Mandatory rather than voluntary gene-synthesis screening, with enforcement. Coordinated AI-and-biology dual-use review across major frontier labs. International agreement on what categories of research are too dangerous for open publication. Sustained funding for the technical-policy bridge, where most of the work has been done by a small number of NGOs (Nuclear Threat Initiative, Center for Health Security at Johns Hopkins, Council on Strategic Risks) on minimal budgets.
What would meaningful reform look like. A package would probably include: a Biological Weapons Convention protocol with at least basic transparency provisions; mandatory gene-synthesis screening with international coordination; pre-publication review of certain categories of dual-use research; AI-bio specific oversight of frontier model deployment with biosecurity-relevant capabilities; substantially increased funding for the small ecosystem of NGOs and academic centres doing the technical-policy work. The political coalition for any of this is fragmented; most of the named individuals listed below have argued for parts of it without producing institutional change.
The China question. Chinese biotechnology has expanded rapidly. Beijing has invested substantially in synthetic biology research, gene-editing capability (the 2018 He Jiankui CRISPR-baby case happened under Chinese oversight that subsequently changed), and biomanufacturing. Western policy debate has often treated Chinese capability as a national-security threat first; the parallel reality is that China's domestic biosafety regulation has tightened in some specific ways since 2018 and is in some respects ahead of Western frameworks. International cooperation on dual-use governance is harder when the geopolitical relationship is adversarial, but it is exactly the kind of issue where cooperation matters most.
The democratization curve
The most underweighted long-run trend in synthetic biology is the falling skill-and-cost barrier to doing significant work. This is mostly a positive story; it is also part of the dual-use picture.
Education. The iGEM (international Genetically Engineered Machine) competition has trained tens of thousands of undergraduate students in synthetic biology since 2003. Standard molecular biology techniques are now taught at the high-school level in some districts. Online courses, free textbooks, and open-source protocols have lowered the entry barrier to the level where motivated amateurs can do real experiments.
DIY biology. Community labs (Genspace in New York, BioCurious in Sunnyvale, La Paillasse in Paris, several dozen others) provide shared lab space for non-affiliated researchers. The DIY-bio movement has been mostly productive (educational projects, citizen-science, open-source instrument development) and self-policing (community norms strongly discourage dangerous work). Whether it stays that way as capability grows is genuinely uncertain.
Lab-in-a-box equipment. The cost and footprint of basic molecular biology equipment has fallen substantially. Open-source thermal cyclers, centrifuges, and DNA-extraction kits enable home or small-lab work that was infeasible a decade ago. Specific equipment for more dangerous work (high-containment laboratories, advanced cell-line manipulation) remains restricted by cost and facility requirements, but the line keeps shifting.
AI as accelerant. Large language models trained on biological literature and AI tools designed for protein engineering have meaningfully lowered the expertise required to do certain kinds of biological work. The 2024 paper by Mouton, Lucas, and Guevara at RAND surveyed how much current AI tools could uplift a non-expert attempting to plan a biological attack and concluded that the uplift was real but partial; subsequent analyses with newer models have generally found larger uplift. The trajectory suggests that the AI-accelerated bio-risk picture in 2030 will look meaningfully different from 2026.
The paths from here
Synthetic biology is unusual among the topics on this site because the trajectory range is genuinely wide. The technology will keep advancing; the institutional response could go several different ways.
Continued capability gains, productive uses dominate
The most likely default. Cheaper sequencing, more clinical CRISPR therapies, more designed proteins, more biomanufacturing, mostly without major incidents. Public attention stays low. Governance gaps persist but do not produce visible failures. Productive uses outpace harm-uses by a wide margin, as has been true for the last two decades.
Will it happen? Probably the base case for the next five to ten years. The dual-use risks are real but most actors who could cause significant harm are also constrained by detection, consequences, and the technical difficulty of weaponising biology effectively at scale.
A specific dual-use incident reshapes oversight
A confirmed lab leak, demonstrated misuse of AI for pathogen design, or smaller-scale biological attack produces sustained political attention. Specific reforms (mandatory gene-synthesis screening, pre-publication review for high-risk research, BWC verification protocol, AI-bio oversight) become globally coordinated. The post-incident regulatory wave is not always well-designed but it would substantially close the current governance gap.
Will it happen? Possible at any time, more likely as capability spreads. The political economy of preventive regulation is weak; the political economy of post-incident regulation is strong. The reforms that would actually help are largely identified; what is missing is the political moment to enact them.
AI-bio acceleration meaningfully changes the threat picture
Frontier AI models trained on biological literature and protein-design tools become substantially better at uplifting non-experts attempting biology work. The capability gap between trained scientists and motivated amateurs narrows. The threat surface for both legitimate and malicious work expands faster than oversight. The frontier-AI labs (Anthropic, OpenAI, Google DeepMind) implement biosecurity-specific safeguards on their models with varying rigour. The picture in 2030 looks meaningfully different from 2026.
Will it happen? Partly already happening. The trajectory of AI capability suggests that the bio-uplift question becomes more pressing each year. Whether governance keeps pace depends on choices being made now in a small number of labs and regulatory bodies.
Therapy access becomes the dominant equity issue
As CRISPR therapies for rare and common diseases multiply, the price tags (currently in the millions per patient for the most advanced therapies) become a central political issue. Insurance systems, national health services, and global access frameworks struggle to absorb the cost. The tension between extraordinarily effective treatments and the question of who gets them shapes a substantial fraction of the broader healthcare conversation. Synthetic biology becomes politically salient mostly through its therapy-cost angle rather than its biosafety angle.
Will it happen? Already happening for the first wave of approvals. The pricing structure and access debates are being formed now. The shape of that conversation matters for whether the technology delivers on its broader promise.
Biomanufacturing produces a quiet industrial transition
Engineered microbes and cell lines start to displace petrochemical-based production for materials, chemicals, fuels, and food at substantial scale. Lab-grown meat reaches commodity price parity for some products. Microbially-produced fragrances, dyes, and structural proteins replace traditional sources. The environmental and supply-chain implications are large; the political attention is currently low. This is a slower-moving but possibly larger long-run story than the therapy or biosecurity angles.
Will it happen? Slowly but surely. Capital investment in biomanufacturing has been substantial since 2020. Scale-up challenges are real; cost-competitive parity in commodity products is harder than in specialty products. The transition will probably be visible in the next decade.
A serious BWC reform negotiation succeeds
The 2026 BWC Review Conference or a successor produces a verification protocol with at least basic transparency provisions - mandatory declarations of dual-use research above some threshold, voluntary inspections, an organisation with technical authority. This would not solve the problem but would meaningfully change the structural picture. The political will required is currently absent; specific events could change that.
Will it happen? Low probability without an incident or a large geopolitical shift. The bilateral US-China relationship in particular is the binding constraint, with neither side currently willing to accept reciprocal verification. Reform is more often produced by crisis than by preparation.
Restriction overshoots and slows productive work
An incident or political wave produces blanket restrictions on synthetic biology research, AI-bio tools, or specific therapy categories. The restrictions are politically rather than technically calibrated and sweep up legitimate work. Productive medical and industrial uses slow down materially. The dual-use risks shift to less-regulated jurisdictions rather than reducing in absolute terms. The net effect on safety is contested.
Will it happen? Plausible if a serious incident occurs without prior frameworks for proportionate response. The pattern in adjacent areas (post-9/11 security restrictions on legitimate research, post-COVID gain-of-function moratorium) suggests overshoot is a real possibility.
The realistic forecast is, as usual, a mix. Continued capability gains and mostly-productive uses are the base case. A specific dual-use incident at some point in the next decade would be unsurprising. AI-bio acceleration is happening regardless. Therapy-access politics are forming now. Biomanufacturing is advancing slowly. Serious international reform is unlikely without an event; overshooting restrictions are a real risk if there is one.
Where serious analysts disagree
Synthetic biology has produced a small but unusually thoughtful debate among researchers, biosecurity analysts, and policymakers. The named voices below are worth reading directly.
The capability is genuinely transformative; the risks are manageable with current institutions
The track record of synthetic biology over twenty years has been overwhelmingly productive. The dual-use scenarios have not materialised at significant scale despite repeated predictions. Existing institutions (research-ethics committees, voluntary screening, professional norms) have done more work than critics acknowledge. The right response is sustained investment in productive applications and incremental strengthening of existing frameworks rather than dramatic restructuring.
Held by: George Church (Harvard), parts of the synthetic-biology research community, the Industrial Biotechnology Innovation Centre community. The case has empirical support in the actual track record; it depends on continuing patterns rather than projections.
The current trajectory is dangerous and requires substantially stronger oversight
The combination of falling barriers (cheaper synthesis, AI uplift, democratised education) and rising capability (designed proteins, AI-aided pathogen engineering) is producing a risk surface that current institutions cannot manage. The COVID experience and the Urbina toxin paper demonstrated that the gap between productive and harmful use can be small. Mandatory gene-synthesis screening, BWC verification, AI-bio oversight, and pre-publication review for high-risk research are needed soon, before an incident makes them inevitable but late.
Held by: Kevin Esvelt (MIT, gene-drive researcher and biosafety advocate), Beth Cameron and Jaime Yassif (Nuclear Threat Initiative), Tom Inglesby (Johns Hopkins Center for Health Security), Greg Lewis and Toby Ord (Oxford, existential-risk framing). The case has empirical support; the institutional traction varies.
AI-bio is the crux; existing biosecurity frameworks were not designed for this
The traditional biosecurity framework assumed that significant biological capability required years of training and institutional support. AI tools are eroding that assumption faster than the framework can adapt. Frontier AI labs need biosecurity-specific safeguards on their models. Open-weights models that include biological-engineering capability are particularly difficult; once released they cannot be recalled. The crux of biosecurity policy in the next five years is at the intersection of AI and biology specifically.
Held by: the AI safety community (Anthropic's biosecurity team, parts of OpenAI's preparedness work, the SecureBio initiative), parts of the RAND biosecurity work (Mouton et al. 2024), the Bipartisan Commission on Biodefense. The case is gaining institutional traction; the implementation is uneven.
Democratization is mostly a positive story
The DIY-bio community has been overwhelmingly productive and self-policing. Educational expansion has produced a generation of researchers who understand both the capability and the responsibilities. Restricting access to the field would primarily damage productive work, since serious bioweapons capability requires resources well beyond what amateurs can muster. The right response to democratization is engagement and education rather than gatekeeping.
Held by: the DIY-bio community (Genspace, BioCurious leadership), the iGEM Foundation, parts of the open-science movement. The case rests on the empirical track record of the community.
The ethics of human germline editing is its own crisis
The 2018 He Jiankui case in China, where edited embryos were used to produce live births, was widely condemned but did not produce durable international restrictions. The technical capability for heritable human genetic modification exists; the ethical and regulatory framework is fragmented across countries. Whether and how this is governed will shape the trajectory of the field substantially. The biosecurity community has paid less attention to this than to the dual-use research debate, but it is at least as consequential.
Held by: Marcy Darnovsky (Center for Genetics and Society), Hank Greely (Stanford bioethics), parts of the National Academies' International Commission on the Clinical Use of Human Germline Genome Editing. The case is contested between bioethics traditions but the empirical question - what is being done now and where - is increasingly important.
None of these readings is fully right or wrong. What can be said from the available evidence: the capability is genuinely transformative, productive uses have dominated and probably will continue to dominate, current institutions are inadequate for the next decade's risk surface particularly at the AI-bio intersection, dramatic restriction would damage productive work without proportionate safety gain, and the right response involves targeted reforms that the political economy has not yet enabled. The next major incident - whatever it is - will substantially shape which of these readings prevails.
What this means for you
If you or someone close to you has a genetic disease
The therapy landscape has changed substantially since 2020 and is changing faster now. Diseases that were untreatable a decade ago may have approved or trial-stage gene therapies. Specific patient-advocacy organisations (the Sickle Cell Disease Association, Cystic Fibrosis Foundation, dozens of others) track approvals and trials at much finer resolution than mainstream coverage. If you have not checked recently, check.
If you read coverage of biotech
Mainstream coverage tends to oscillate between "miracle cure" and "Frankenstein" framings, both of which are usually wrong about specific technologies. Specialised sources (STAT News for medical biotech, Endpoints for industry, the Bulletin of the Atomic Scientists for biosecurity, the Council on Strategic Risks for dual-use) consistently outperform mainstream coverage on accuracy. Original research papers are more accessible than non-scientists usually realise; the abstract and discussion sections are typically readable.
If you vote on biotechnology or biosecurity policy
The unglamorous questions matter most. Funding for the small ecosystem of biosecurity NGOs and academic centres. Mandatory gene-synthesis screening with enforcement. AI-bio oversight at frontier labs. BWC verification position. Therapy-access frameworks for hereditary disease treatments. Dual-use research review structure. These are not headline issues; they are where the actual trajectory is being shaped. Politicians who engage them substantively are doing some of the most consequential work; politicians who treat biotechnology as ordinary partisan material usually have not done the work.
If you are considering a career in the field
The opportunity space is unusually large. Computational protein design, AI-bio interface, biomanufacturing scale-up, regulatory science, and biosecurity-policy work are all under-staffed relative to demand. The career path through standard biology graduate training is well-mapped; the careers at the bridges (computational + biology, science + policy, biology + safety engineering) are less well-mapped and probably more impactful. Several specialised graduate programs and fellowships have appeared in recent years.
If you are anxious about the dual-use risk
The honest picture is that the risk is real but is also one of many. Pandemic preparedness, AI safety, nuclear arms control, and climate adaptation all matter and all are under-funded relative to expected harm. Personal action that helps: supporting NGOs working on the structural questions (Nuclear Threat Initiative, Center for Health Security, Council on Strategic Risks); advocating with elected representatives for the unglamorous reforms; engaging substantively rather than from alarm. Catastrophising rarely produces good policy; sustained attention to the structural questions sometimes does.
