Anthropic reports attempts to misuse Claude for cyberattacks and biological research
In the UAE, The National reported that Anthropic had published a report on several cases of attempted misuse of its artificial intelligence models. According to the company, Claude was allegedly used for cyberattacks, influence operations, surveillance, biological research, and weapons development.
Attempts to misuse models
Anthropic said it had disrupted all operations mentioned in the report and was using the investigation findings to strengthen protections for its services. The company also said it was blocking accounts linked to the detected activity and introducing new measures to detect and prevent misuse.
One example involved a request to Claude for help preparing a grant application for research into the chikungunya virus. This virus is transmitted by mosquitoes; it causes debilitating symptoms, including severe pain and fever, which can last for weeks or months. The report also states that there are no licensed treatments for the virus.
More current news is available on the UA.News Telegram channel Telegram.
Anthropic said that the institutional affiliation connected to the grant application raised suspicions. The company did not specify which institution was involved. The report describes such attempts as actions that could potentially facilitate the development of biological weapons.
Safety and scientific use
The company emphasized that it seeks to preserve the usefulness of its models for scientific research. Anthropic predicts that artificial intelligence could transform biological science and accelerate the development of new treatments, while acknowledging the difficulty of establishing safeguards that adequately account for context.
The report was published amid discussions in the United States about the risks posed by powerful AI models. Some members of the US Congress introduced bills to halt or restrict research in artificial intelligence, but none of these proposals became law.
