Friday, October 30th, 2026
Sign in to LinkedIn, then select Subscribe.
FOUNDATIONAL UPDATES
Anthropic shares early results from six drug development partners, including Claude reviewing GMP batch records and drafting IND sections
On October 27, Anthropic posted an update on its work with six organizations developing medicines: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. The most hands-on result came from Enveda, where manufacturing chemists had been spending up to 75% of their time on paperwork. Enveda connected its manufacturing documents to Claude and had it flag possible issues. Across 1,061 pages, Claude caught all 11 issues human reviewers had found and flagged seven they'd missed. Review time fell 89%, from more than 25 hours to under three. Enveda's own release says its Quality team still does a final review of every record before any batch is released.
Claude also wrote first drafts of two regulatory documents, including big portions of Module 2 of an IND. One took four hours versus about three months by hand. At BMS, Claude is being used to assess plans for an early-stage drug in clinical trials and estimate the chance those trials succeed, with more work planned for 2027. Michael Pourdehnad, M.D., BMS SVP and head of early clinical development for hematology, oncology and cell therapy, said a single prompt produced "a robust initial analysis" that would otherwise take days. Anthropic also acquired Mithril Medicine, a one-person company working on two rare forms of muscular dystrophy, and expects a drug candidate could be ready for human trials as soon as late 2028. Anthropic says the work is still mostly preclinical.
Why it matters for pharma: Batch record review and IND writing sit right in the regulated part of the job, and now there are hard numbers on both. Two things to keep straight. This was an internal evaluation on one company's documents, not a validated system. And the human Quality release didn't go away. If you pilot something similar, write down how you'll handle what the model flags that your reviewers didn't, and what it misses. That's the first thing an inspector will ask about.
Source: Anthropic, "Early progress on accelerating drug development with Claude" https://www.anthropic.com/research/early-progress-drug-development
Google DeepMind and Springer Nature build an AI assistant for journal peer reviewers
On October 26, Springer Nature and Google DeepMind announced a research partnership to co-design an AI assistant for peer reviewers, built on DeepMind's Gemini models. The prototype is in testing now. It helps reviewers work through long manuscripts, charts and figures and points out spots that may need a closer look, so reviewers can spend more time on questions that need a specialist. It runs only inside Springer Nature's private Gemini Enterprise environment, so manuscripts used to build and test it stay inside the publisher's systems. Saskia Steinacker, Springer Nature's chief product officer, said the company wants to be "clear about where responsibility remains."
Why it matters for pharma: Your trial results papers will start getting read by an AI assistant before, or alongside, a human reviewer. So medical writing and publications teams should tighten the match between the manuscript, the CSR and the registry posting. A tool that reads every table won't miss an inconsistency a tired reviewer might. The setup is worth copying too. Confidential documents stayed inside the publisher's own environment, and the human kept the decision. Sponsors putting AI on unpublished data should ask vendors for the same.
Source: Springer Nature press release https://group.springernature.com/us/group/media/press-releases/research-partnership-with-googledeepmind-to-support-peer-review/53891572
DRUG DEVELOPMENT & AI
AstraZeneca and Tremont AI team up on AI for preclinical safety pathology
On October 27, Boston-based Tremont AI announced a research collaboration with AstraZeneca to build and test AI for faster safety screening in preclinical toxicologic pathology studies. In these studies, pathologists review large numbers of tissue images from animals given a drug candidate. They look for drug-related changes and separate them from normal biological variation. Those calls help set a compound's safety profile and the dose that can be given safely. Tremont says the work today relies on expert reads that vary between reviewers, and that its models can make it measurable and repeatable. Tremont is a spinout of the Mahmood Lab at Mass General Brigham. No financial terms were disclosed.
Why it matters: Tox pathology feeds the starting dose in a first-in-human protocol, so this is about as close to the IND as preclinical AI gets. If a model ends up triaging slides in GLP studies, the study record has to show what the model looked at, what it flagged, and what the pathologist decided. Development teams should ask their CROs now whether AI-assisted reads are coming and how they'll be documented in the final report.
Source: Tremont AI https://tremont.ai/news/tremont-ai-announces-research-collaboration/
Boehringer Ingelheim and Microsoft explore AI "digital twins" for trial design
Boehringer Ingelheim and Microsoft have started a research collaboration on digital patient twins. These are computer models of individual patients that use clinical and other data to predict how a disease might progress under different scenarios. The companies announced it October 22, and Fierce Biotech reported details October 26. A Boehringer spokesperson told Fierce that "the aim is not to mimic or replace an entire clinical trial." One approach they plan to test is using twins to build synthetic or hybrid control arms, which could cut the number of patients needed in a standard control group. Work starts in lung health. The spokesperson added that "any application would require rigorous validation and regulatory acceptance."
Why it matters: Smaller control arms are the prize here, and also the hardest thing to get past a regulator. Anyone heading down this road should agree the model's intended use and its validation plan with FDA and EMA well before it touches a pivotal design. And keep the data lineage clean from day one. Any model built on pooled trial and real-world data will get asked where every record came from.
Source: Fierce Biotech https://www.fiercebiotech.com/cro/microsoft-bi-team-launch-digital-patient-twins-trial-design-improvements
REGULATORY, POLICY & INDUSTRY MOVES
UK launches an AI Biosecurity Lab backed by OpenAI and Anthropic, with a tie to MHRA on AI-designed medicines
On October 24, the UK Cabinet Office announced the AI Biosecurity Lab, which it calls the world's first dedicated AI biosecurity organization. It has £100 million in government funding and up to £200 million in company commitments. OpenAI will provide £100 million in compute and model credits plus trusted access to its most capable relevant models. Anthropic will provide up to £100 million in model credits, trusted model access and technical support. Google DeepMind and UK company Basecamp Research will partner on research. The lab will run as an independent non-profit and will focus first on earlier detection of biological threats and faster development of vaccines and treatments. It'll work with MHRA "to prepare for health emergencies and inform the regulation of AI-designed medicines."
Why it matters: That last line is the one to watch. A government lab with deep access to frontier models will be helping MHRA think about how to regulate medicines AI designed. Also notice the "trusted access" setup. The most capable biology tools are increasingly gated behind vetting programs, and that will decide which teams can use which models on real programs. If your company wants in, expect to show governance and controls to get it.
Source: GOV.UK https://www.gov.uk/government/news/introducing-the-ai-biosecurity-lab
Senate health chair Bill Cassidy floats a draft framework for AI in health care
On October 29, Sen. Bill Cassidy (R-La.), the outgoing chair of the Senate HELP Committee, released a discussion draft meant to promote access to AI health technology while setting guardrails. According to Bloomberg Government, the draft would require HHS to develop use cases for AI. "Implementing AI in health care must be done responsibly and to the benefit of patients, innovators, and consumers," Cassidy said. The public coverage so far doesn't go past the summary, so it's not yet clear what the draft means for FDA or for drug and biologic sponsors.
Why it matters: It's a discussion draft from a chair who's leaving, so nothing changes tomorrow. Still, regulatory policy teams should pull the full text and see whether it touches FDA's authority over AI used in drug development or only care delivery. Comments on discussion drafts are how industry gets heard early.
Source: Bloomberg Government https://news.bgov.com/bloomberg-government-news/cassidy-releases-draft-proposal-to-regulate-ai-in-healthcare
Originally published on LinkedIn. Read the original
