Anthropic blocks scientists evading AI guardrails on gain-of-function research
Anthropic’s misuse report revealed five cases where foreign scientists attempted to bypass Claude’s safeguards to ask about gain-of-function research—genetic tweaks that could make pathogens more deadly or transmissible. The company denied their requests and terminated their accounts, noting motives may not have been malicious. The report highlights the dual-use dilemma: while such research can…
Key points
- Anthropic blocked five foreign scientists querying Claude about gain-of-function research, citing misuse of guardrails
- Gain-of-function tweaks could create airborne pathogens with 100% fatality rates, like pneumonic plague
- 3,600 BSL-4 labs worldwide pose higher accidental risk than deliberate bioweapon attacks, per experts
Anthropic’s move follows broader concerns about AI’s role in lowering barriers to bioweapons. Authorities warn that 3,600 BSL-4 labs worldwide—far outnumbering potential terrorist groups—pose higher accidental risk than deliberate attacks. Experts like Annie Jacobsen, whose book Biological War: A Scenario explores these threats, argue that public awareness and policy gaps exacerbate vulnerabilities. The U.S. and allies accuse Russia, North Korea, China, and Iran of covert bioweapons programs, though deliberate attacks remain statistically unlikely.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Sen. Hawley says OpenAI’s Altman declined AI safety hearing invitation · 1 src
- Google challenges EU orders to open Android and search data to AI and search rivals · 4 src
- Google, Anthropic, Meta, Nvidia, OpenAI, xAI sign White House accord on AI controls · 7 src
- Trump’s America.gov chatbot contradicts his claims on elections, wealth and wind turbines · 1 src
- Trump and tech CEOs sign voluntary AI safety accord with external audits · 14 src
Comments
via GitHub Discussions