Anthropic, the company behind the Claude family of artificial intelligence models, said it identified and blocked efforts to use its systems to develop biological weapons. The disclosure arrived alongside a broader account from the company of attempts by outside actors to misuse its technology, and it landed in the middle of a wider debate over how quickly powerful AI systems should be built and deployed.

Anthropic said governments and other groups have sought to turn its models toward tasks including automated intelligence gathering. The company reported that actors it associated with China and Russia attempted to weaponize its tools. Anthropic said it moved to detect and stop those efforts.

At the same time, the company has been expanding cooperation with government bodies focused on security. Anthropic granted the European Union's cybersecurity agency access to one of its AI models, part of a pattern in which the firm has presented its technology as both a potential risk and a tool for defending against misuse.

Running parallel to these disclosures is a debate inside the industry about the pace of development. A researcher resigned from Anthropic with a public warning about the dangers of continued rapid AI advancement. That departure came amid a broader set of calls from researchers connected to Anthropic and OpenAI for a slowdown, citing concerns about safety and oversight.

Some of those warnings have been stark. One Anthropic researcher put the probability that AI could "kill all humans" within the next decade at more than 10 percent, a figure that reflects the language some in the field use to describe worst-case scenarios. Such estimates are contested, and not everyone in the industry accepts either the numbers or the framing behind them.

That skepticism has its own voice. National Review published a critique of what it termed AI "doomer" culture, arguing that dire predictions carry little accountability when they do not come to pass. The piece reflected a strand of commentary that questions whether extinction-level warnings are useful or verifiable.

Taken together, the coverage points to a company navigating two roles at once. Anthropic is describing threats it says it has blocked and partnerships it has formed to counter misuse, while some of its own current and former researchers press the argument that the technology is advancing faster than safeguards can keep up. The disagreement over how seriously to weigh catastrophic risk remains unsettled, both inside Anthropic and across the broader field.

References

1. 1.Anthropic — disclosures on blocking biological weapons attempts and misuse by actors it associated with China and Russia
2. 2.Axios — Anthropic granting the EU cybersecurity agency access to an AI model
3. 3.Fox News — Anthropic researcher's estimate of more than 10 percent probability AI could 'kill all humans' within a decade
4. 4.National Review — critique of AI 'doomer' culture and the accountability of dire predictions
5. 5.Public statements — researcher's resignation from Anthropic and calls for a development slowdown