Healthcare
AI Discovers CRISPR-Like Enzyme System: Could Generative Biology Become the Next Trillion-Dollar Industry?
26 Oct 2026
Anthropic's Claude AI agents have identified a previously uncharacterised biological system with similarities to CRISPR, while researchers using the Evo 2 model have successfully designed bacteriophages capable of killing bacteria. Together, these developments point towards a new era of AI-assisted biological discovery, with major implications for biotechnology, pharmaceutical research and investment.
Artificial intelligence is moving beyond generating text, images and software code into an increasingly complex field: the discovery and design of biological systems.
In September 2026, Anthropic announced that approximately 950 Claude AI agents had spent 21 hours examining large biological datasets, identifying an unusual enzyme-associated system with structural similarities to CRISPR, the technology that transformed genetic engineering.
The discovery, named Array-Associated Reverse Transcriptases (ART), demonstrates how autonomous AI systems can assist researchers in analysing enormous volumes of genetic information and identifying previously overlooked biological patterns.
Around the same period, researchers at Stanford University reported another development involving Evo 2, a generative AI model developed with the Arc Institute. The researchers used AI-generated genetic sequences to create bacteriophages capable of killing Escherichia coli (E. coli) bacteria in laboratory experiments.
Writing in a September 29, 2026 Medium article, technology commentator Mohamed Abdelmenem argues that these developments could signal the emergence of a new economic sector built around generative biology, automated laboratories and AI-driven pharmaceutical discovery.
His central prediction is that the combination of autonomous AI agents and biological engineering could create a major new investment opportunity by 2028. However, the commercial scale and timeline remain speculative.
How 950 Claude AI Agents Identified a New Biological System
Anthropic's discovery began with a broad research objective: examine large collections of DNA sequences to identify potentially interesting reverse transcriptase systems.
Reverse transcriptases are enzymes that convert RNA into DNA. They play important roles in biological processes and are already used in research and biotechnology.
According to Anthropic, approximately 950 AI agents worked through the dataset over 21 hours, using around 210 million tokens of computation.
The agents gathered and analysed more than 200,000 reverse transcriptase sequences, investigated thousands of candidate systems and narrowed their findings for closer examination.
During the process, one agent identified an unusual arrangement of repeating DNA sequences located near a reverse transcriptase gene.
The pattern resembled aspects of CRISPR systems, prompting further investigation by human researchers.
Anthropic subsequently named the system ART.
The finding is notable because the individual reverse transcriptase had been identified previously, but the broader arrangement of associated genetic elements had not been characterised in the same way.
Importantly, ART has not been demonstrated to function as a CRISPR gene-editing tool. Its precise biological role remains unknown, and further experimental research is required to establish its potential applications.
Stanford's Evo 2 Demonstrates AI-Designed Biological Systems in the Laboratory
A second development provides evidence that generative AI can contribute not only to biological discovery but also to the design of functional genetic systems.
Researchers at Stanford University used Evo 2, a genomic foundation model developed in collaboration with the Arc Institute, to generate novel bacteriophage genomes.
Bacteriophages are viruses that infect bacteria, making them an area of interest in research into alternative treatments for bacterial infections.
Starting with genetic information from an existing bacteriophage known as ΦX174, the researchers generated new DNA sequences using the AI model.
These sequences were subsequently selected, synthesised and tested in a physical laboratory.
The team produced and evaluated nearly 300 phages, eventually identifying 16 particularly effective candidates against E. coli.
The experiments demonstrated that AI-generated genetic designs could produce functional biological results rather than remaining entirely computational predictions.
While promising, these experiments do not establish that AI-designed phages are ready for clinical use or that traditional antibiotic development can be replaced.
They do, however, demonstrate a growing connection between computational biological design and laboratory experimentation.
The Convergence of AI Agents and Generative Biology
Abdelmenem argues that the significance of these discoveries lies in the combination of two emerging capabilities.
The first is autonomous research, where AI agents can examine scientific data, generate hypotheses and identify promising areas for further investigation.
The second is generative biological modelling, where AI systems can propose new DNA, RNA or protein sequences for laboratory testing.
Together, these technologies could create a more integrated discovery process, beginning with computational analysis and progressing towards physical experimentation.
Instead of relying exclusively on researchers to identify and investigate every potential biological candidate manually, AI systems could help narrow the range of possibilities requiring laboratory evaluation.
This does not eliminate the importance of scientific expertise. Human researchers remain essential for establishing experimental objectives, assessing biological relevance, evaluating safety and interpreting results.
Nevertheless, greater automation could influence how biotechnology companies organise early-stage research and allocate resources.
Why Physical Laboratories Remain the Biggest Challenge
Although artificial intelligence can process biological information at remarkable speed, experimental biology remains constrained by physical processes.
DNA synthesis, cell cultivation, molecular testing, equipment availability and experimental validation all require time and resources.
Abdelmenem describes this challenge as the wet-lab wall: the gap between rapid computational discovery and slower physical experimentation.
An AI model may generate thousands of promising biological sequences, but researchers still need to determine which candidates function as predicted.
Laboratory testing may reveal problems that are not apparent in computational models, including unexpected biological interactions, poor performance or safety concerns.
This creates demand for systems capable of connecting digital research tools with automated laboratory equipment.
Robotic experimentation platforms, laboratory management software and integrated biological data systems could help improve this process, although their economic benefits will depend on reliability, validation requirements and actual research outcomes.
Could Generative Biology Transform Pharmaceutical Research?
For pharmaceutical companies, AI-assisted biology could influence the early stages of drug discovery and development.
Potential applications include identifying new biological targets, generating molecular candidates and evaluating biological mechanisms before committing resources to extensive laboratory testing.
Abdelmenem argues that these capabilities could alter the economics of research by reducing the manual effort required for some discovery activities.
However, the claim that AI computation has already made traditional pharmaceutical research economically obsolete is not supported by the discoveries discussed.
Computational identification is only one stage of pharmaceutical development. Experimental validation, toxicology, clinical trials, manufacturing and regulatory review remain essential.
Consequently, faster discovery does not automatically translate into faster drug approvals or lower total development costs.
The longer-term opportunity may lie in integrating AI-assisted discovery with improved experimental infrastructure, allowing researchers to investigate more candidates while maintaining rigorous scientific standards.
The Investment Opportunity in Automated Biological Infrastructure
Abdelmenem's investment thesis centres on the infrastructure required to connect generative AI with physical biological research.
He identifies automated laboratory systems, biological data tools and simulation platforms as potential areas of commercial opportunity.
| Technology area | Potential role in generative biology |
|---|---|
| AI genomic models | Generating and analysing biological sequences |
| Autonomous research agents | Identifying candidates and developing research hypotheses |
| Laboratory robotics | Automating selected experimental procedures |
| DNA synthesis infrastructure | Producing genetic material for research |
| Biological data platforms | Connecting research software with experimental systems |
| Digital biological simulations | Evaluating candidate designs computationally |
These areas could benefit from increasing demand for AI-assisted biotechnology, but their future profitability is uncertain.
Commercial success will depend on factors including research reliability, regulatory requirements, development costs, customer adoption and the ability to demonstrate measurable improvements over existing methods.
Abdelmenem argues that investors should pay particular attention to companies building automated laboratory infrastructure rather than focusing exclusively on general-purpose AI applications.
His assessment reflects a forward-looking investment opinion, not a demonstrated market valuation or guaranteed industry trajectory.
What Could Happen by 2028?
One of the article's most ambitious predictions is that an autonomously AI-designed drug could enter human clinical trials by the first quarter of 2028.
Abdelmenem presents this as an 18-month period during which the capabilities of generative biology may become clearer.
The prediction depends on substantial progress in connecting computational models with physical laboratories and demonstrating that AI-generated biological candidates can satisfy rigorous testing requirements.
The author also identifies three developments worth monitoring: improved integration between AI software and laboratory equipment, regulation of openly available genomic AI models, and regulatory submissions involving increasingly autonomous drug-development workflows.
Biosecurity will be another important consideration. Advanced models capable of proposing biological sequences may offer valuable research opportunities while also raising questions about potentially harmful applications.
Developing appropriate safeguards, oversight and evaluation systems will therefore be important as these technologies become more capable.
There is currently no established basis for treating the first quarter of 2028 as a confirmed milestone for an autonomously AI-designed drug entering human trials.
A New Phase in AI-Driven Scientific Discovery
The developments involving Anthropic and Stanford illustrate how artificial intelligence is beginning to participate more directly in scientific research.
Anthropic's ART discovery demonstrates the ability of AI agents to identify unusual patterns in large biological datasets, while Stanford's Evo 2 experiments show how generated genetic sequences can be tested for biological function.
The next challenge is translating these capabilities into reliable, repeatable and economically useful research workflows.
For biotechnology companies, the potential benefits extend beyond accelerating individual discoveries. AI could change how researchers prioritise experiments, coordinate laboratory work and evaluate new biological possibilities.
For investors, however, the distinction between technological potential and commercial maturity remains essential.
Generative biology may become an increasingly important industry, but predictions of trillion-dollar valuations and fully autonomous drug development still require substantial evidence.
The most significant development may not be that artificial intelligence has replaced biological researchers, but that it is becoming a new tool for scientific discovery, extending what human researchers can investigate and test.
Sources
- Mohamed Abdelmenem, AI Just Discovered a CRISPR-Like Enzyme. Here Is What Happens by 2028, Medium, 29 September 2026. Original commentary supplied for adaptation. Author: https://medium.com/@mohamed-abdelmenem
- Anthropic, Claude Discovers a Novel Enzyme System, September 2026: https://www.anthropic.com/research/claude-discovers-novel-enzyme-system
- Stanford University, AI Designs a Novel E. coli Killer, 6 August 2026: https://news.stanford.edu/stories/2026/08/evo-2-ai-tool-e-coli-killer-bacteriophages
- IBM Think, What Anthropic's AI Biology Experiment Actually Showed, 5 October 2026: https://www.ibm.com/think/news/what-anthropics-ai-biology-experiment-showed
