Anthropic announced Thursday that it has intercepted multiple misuse attempts involving its artificial‑intelligence models. The company says bad actors tried to employ its technology for cyber‑attacks, surveillance and even research that could support the creation of biological weapons.
New safeguards added to newer models
As AI systems become more capable, the company warned that sophisticated threats no longer require large teams of experts. Lone individuals can now generate dangerous code or instructions that were impossible a year ago. In response, Anthropic has strengthened safeguards in its latest Claude Fable 5 and Mythos‑class models to block queries related to dual‑use biological research.
Examples of blocked misuse
The third misuse report released by Anthropic, covering December 2025 through August 2026, includes several notable cases. One involved a request for Claude’s assistance in drafting a grant proposal for gain‑of‑function research on the chikungunya virus – a mosquito‑borne illness that can cause severe pain and fever. While such research could help develop vaccines, the company noted it could also be used to make the pathogen more dangerous.
Another case described an “industrial‑scale, covert campaign” to extract capabilities from an older model and replicate them in a new one without authorization. The company said its older Claude Opus 4 and Claude Sonnet 4.5 models were below the threshold for assisting sophisticated biological research, but newer models now pose a higher risk.
Broader threat landscape
Anthropic also identified influence‑operation activity, noting groups that created hundreds of fake social‑media accounts to amplify political messages across Russia, Iran, Turkey, the Persian Gulf, South Asia, Africa and Europe. The firm warned that AI platforms may detect such operations only after they have begun to spread.
Company response and industry call‑to‑action
Anthropic said it has blocked each of the identified malicious activities, shared findings with government authorities and industry partners, and used the experience to further tighten its safeguards. “We believe we have a responsibility to disclose malicious misuse of our services,” the company wrote. “As models become increasingly capable, their risks will increase unless AI developers and society’s defenders act to make them safer.”
The report follows the resignation of Anthropic researcher Jacob Coxon, who warned that the race toward self‑improving superintelligence could endanger human life by the end of the decade. Coxon’s concerns echo broader industry debates about responsible AI development.
Looking ahead
Anthropic plans an initial public offering later this fall and hopes its transparency will encourage other developers to recognize similar misuse patterns. The company urges governments and civil society to gain a clearer view of emerging threats and to strengthen collective defenses against AI‑enabled abuse.
Original reporting: KTBS 3 (Shreveport) — read the source article.
