Chicago – Anthropic, the creator of the Claude family of artificial‑intelligence models, released its third misuse report Thursday, detailing how the company intercepted multiple attempts by bad actors to employ its technology for harmful purposes. The most alarming case involved a request to help draft a grant proposal for gain‑of‑function research on the chikungunya virus, a mosquito‑borne illness that can cause severe pain and fever.
Stronger safeguards for newer models
Anthropic said its older models, such as Claude Opus 4 and Claude Sonnet 4.5, were “well below the threshold” needed to meaningfully assist sophisticated users in dangerous biological research. Consequently, those models received lighter protections focused on preventing novice attempts to recreate known bioweapons.
In contrast, the company’s latest offerings – Claude Fable 5 and the Mythos‑class series – now carry “stronger safeguards that restrict access to a wide range of dual‑use biological research queries.” This change follows the discovery that even powerful AI can be leveraged to accelerate complex scientific tasks, including those that could be repurposed for weaponization.
Examples of blocked misuse
The report cites nine distinct incidents, ranging from spyware vendors and politically motivated individuals to state‑sponsored propaganda outfits. One highlighted episode involved an unnamed actor attempting to use Claude to author a grant application aimed at enhancing the transmissibility and immune‑evasion properties of the chikungunya virus. While such research could theoretically aid vaccine development, Anthropic warned it “could also be used to make the pathogen more dangerous.”
Another case described an “industrial‑scale, covert campaign” to extract capabilities from a model and replicate them in another without authorization. Anthropic blocked the attempt and flagged it as a serious breach of its usage policies.
Company response and industry context
Anthropic emphasized its responsibility to disclose malicious misuse and urged both governments and rival AI developers to adopt similar transparency. “As models become increasingly capable, their risks will increase, unless AI developers and society’s defenders act to make them safer,” the company said.
The report arrives amid growing calls for federal regulation of AI. Experts such as John Thickstun, an assistant professor of computer science at Cornell University, argue that private firms are placed in an uncomfortable position when asked to make “value judgments at societal scale without any kind of democratic or deliberative oversight.”
Internal dissent and future outlook
Two days after the report’s release, Anthropic researcher Jacob Coxon announced his resignation, citing concerns that the industry is “racing straight to self‑improving superintelligence and gambling with our lives.” Coxon warned that some colleagues now believe AI could pose an existential threat by the end of the decade.
Anthropic says it has shared the findings with government authorities and industry partners, hoping the information will help others recognize similar patterns and strengthen collective defenses against AI‑enabled threats.
Original reporting: Brookhaven News – ABC7 New York — read the source article.
