Drug development contains many distributed bottlenecks
Anthropic has summarised early work with BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group, along with drug programs it now operates internally. The company argues that AI can assist across target research, molecular design, manufacturing, regulatory preparation and clinical planning rather than solving only one discovery step.
In the BigHat collaboration, a researcher working with Claude proposed antibody changes while an automated laboratory produced and tested about 1,010 variants over roughly ten weeks. The teams also reconsidered misleading stability and aggregation metrics. Anthropic reports that selected variants improved binding, heat stability and resistance to clumping compared with their starting molecules.
Claude participates in design, analysis and operations
Generate Biomedicines used Claude and other agents to propose shorter signalling domains for cell therapies. Scientists reviewed the reasoning and selected 186 designs for initial laboratory testing. Divergence used Claude to explore complex antibody designs for food allergies, but reported that humans remained better at deciding when a result was definitive and when to change research direction.
Anthropic has also acquired Mithril Medicine, whose founder used Claude to coordinate work across eleven external laboratories for rare muscular-dystrophy programs. Early mouse results are described as promising, and the team aims to identify formal drug candidates in 2027. Those timelines are targets, and the work remains well before evidence from human clinical trials.
Preclinical progress is not clinical proof
The company acknowledges that the results are mostly preclinical. Models can accelerate literature review, propose experiments, analyse raw curves and reduce administrative load, but they cannot remove biological uncertainty or replace regulated testing. A design that improves a laboratory metric may still fail on toxicity, manufacturing, efficacy or patient variability.
The work is most credible where model proposals are paired with physical experiments and expert decisions. Researchers should preserve hypotheses, rejected options and protocol changes so positive results can be reproduced. Publication should distinguish what Claude generated, what scientists changed and which measurements were independently verified, especially when commercial partnerships provide credits or engineering support.
What teams should validate before production
For implementation, teams should record the exact model, software, dataset, connector and policy versions behind every result. Representative evaluation must include normal work, boundary cases, deliberate misuse and component failure, with outcomes separated by user role, data sensitivity and consequence. A strong average should never hide a serious failure in a critical path. Changes should pass the same regression set before promotion, and operators need a documented fallback that preserves evidence rather than silently broadening access or lowering quality.
Procurement and planning should account for the human work required to review exceptions, maintain controls and investigate incidents. Published performance or efficiency claims are useful starting points, but production confidence comes from evidence collected in the organisation’s own environment. A named owner, measurable acceptance threshold, cost model and tested shutdown path turn a promising release into a service that can be governed over time. Where decisions affect safety, rights, money or regulated records, accountable professional review remains essential.
Security and operational evidence
Security review should map every identity, credential, network destination, stored artefact and administrative access path involved in the workflow. Logs need enough context to reconstruct who initiated an action, which evidence was consulted, what the system proposed and what ultimately executed. Sensitive prompts and traces should not be retained merely because they are useful for debugging. Retention, redaction and access must follow the underlying data classification, with a practical process for correcting or removing material when policy or law requires it.
Operational monitoring should measure quality and service behaviour together. Latency, throughput and availability matter, but so do unsupported claims, missed constraints, unsafe tool calls and the rate at which people overturn recommendations. Teams should rehearse dependency outages and partial failures before unattended operation. If a model, connector or external service becomes unavailable, the workflow should fail predictably, preserve its state and tell the user what remains unfinished. Clear degradation is safer than an agent improvising around a missing control.
Transparency should scale with adoption
Organisations should explain the scope and limitations of the deployed capability to the people affected by it. Users need to know when automation is involved, how to challenge an outcome and where a human can intervene. That transparency improves feedback and prevents a pilot result from being interpreted as blanket approval for broader autonomy. Expansion should follow demonstrated evidence, not novelty or pressure to match a competitor announcement. The durable advantage comes from dependable integration, careful measurement and learning that survives the next model or product update.
Finally, teams should schedule a review after real usage rather than treating launch as the end of governance. The review should compare expected benefits with observed quality, cost, incidents and user behaviour, then retire workflows that do not justify their complexity. A service that works well can expand through the same evidence-based process. A service that fails should be narrowed or stopped without losing the records needed to understand why. This feedback loop is what turns experimentation into accountable operational capability.