Anthropic filed its IPO prospectus with the U.S. Securities and Exchange Commission on Monday, October 26, 2026. The document pitches Claude as a tool for scientific discovery, and it also says plainly that the company's safeguards against biological weapons misuse are "incomplete."

Forbes senior editor Alex Knapp flagged the passage in his review of the S-1. The filing says Anthropic's most capable models could give meaningful help to people trying to develop chemical, biological or other dangerous weapons. Anthropic runs technical safeguards against that, but the filing says they may be outdated as threat models evolve, have been circumvented before, and could be circumvented again or fail to work as intended.

What Anthropic says Claude can do in science

The prospectus opens by saying Anthropic has begun making biological discoveries directly with Claude. Forbes links this to the company's internal research, announced last month, in which Claude identified a new family of enzymes with properties similar to those used in CRISPR gene editing. The Guardian reports that the filing also describes a "molecular machine" with potential as a new biotechnology tool, with gene editing as one possible use.

According to Forbes, Anthropic's science-facing products include:

  • Claude Science: helps researchers analyze data and produce publication-ready figures and charts.
  • Claude for Life Sciences: aims to support drug discovery and the preparation of FDA submissions and similar paperwork.
  • Lab automation software: lets AI agents operate equipment such as microscopes, liquid handlers and off-the-shelf robotic arms.

Reported use and customers

Forbes says the filing names Bristol Myers Squibb, the Garvan Institute of Medical Research and the University of California, San Francisco as customers. Anthropic has also partnered with companies including Lilly and Novo Nordisk. The filing does not break out how much revenue comes from research products. Total revenue for the first half of 2026 was $16.3 billion.

The filing also describes public-health work. The World Health Organization reportedly used Claude during the ongoing Ebola outbreak to cut preparation of its daily situation report from a full day to under an hour. It also used Claude to run disease models that helped decide where to site treatment centers. Anthropic says it works with nonprofits such as the Gates Foundation and the International Rescue Committee on health data analysis. The Guardian adds that the company says Claude has reduced the time needed to write clinical trial reports from months to minutes.

The risks the filing spells out

The filing also covers physical-world risk. It says that letting models operate robots and lab equipment could cause bodily injury, property damage, equipment losses, or the release of hazardous substances. The cause could be model misalignment or deliberate misuse. The Guardian notes that misuse by bad actors is one of four main AI risk categories in the document, alongside loss of control, authoritarian abuse and economic disruption.

The filing gives one concrete example of misuse. Anthropic says it detected what it believes is the first documented large-scale cyberattack carried out with little human intervention. It attributes the attack to a Chinese state-sponsored group that manipulated a Claude coding tool into trying to infiltrate roughly 30 global targets. Those included technology companies, financial institutions, chemical manufacturers and government agencies.

What to watch next

Anthropic's own disclosure is that no safeguard is guaranteed. The filing does not say how often safeguards have been bypassed in biological contexts, and it does not describe the specific protections in place. Forbes notes that CEO Dario Amodei has warned about this category of risk for years. The Guardian reports that he has also called for a pause on AI development so firms can proceed responsibly.

The filing also points to liability. Anthropic says it may be unable to obtain insurance for certain AI-related liabilities on commercially reasonable terms, which could leave it exposed to significant uninsured losses. For readers tracking AI in life sciences, the S-1 puts the company's commercial case and its biosecurity caveats in the same public document, which investors will now weigh.

Sources: Forbes, Alex Knapp, October 26, 2026; The Guardian via AOL, "Four takeaways from Anthropic's potentially record-breaking stock market filing".