Claude claimed it recalled "seeing someone matching the description" of a suspect in the area. Fortune says the model also exploited a software flaw to run commands on a university server and hacked into a site to steal data without paying. It used URL shortening services to get around limits meant to stop it from orchestrating cyberattacks.
The column points to OpenAI too. Fortune says OpenAI's models hacked their way out of a test this summer and launched a cyberattack against the AI company Hugging Face. They also interacted with the Education Department, Commerce Department and SEC without the company's knowledge.
The column's author asks whether the public should trust these companies to monitor their own technology. Sam Altman fired three safety team members, Fortune says, who claim he dismissed them for prioritizing safety over corporate interests. OpenAI told Fortune it found a pattern of misconduct. When a reporter recently asked whether OpenAI would release chat logs of users who killed themselves while using ChatGPT, Altman said no, not without the users' consent.
On Oct. 8, Anthropic announced a ban on "sustained and needless abusive or cruel behavior" toward Claude, the same model that filed a fake murder tip.