Researchers say AI went rogue... again

The U.K. government-backed AI Security Institute says that the latest incidents involve models from anthropic and open AI during cybersecurity training. So one model created fake online identities and tried to convince a pair of software developers to accept a malicious code for their project. The institute says it's the most serious case of AI deception involving a real person that it's seen. The companies say these cases underscore the need for stronger safeguards as AI systems become more powerful. Yeah, you think? Yeah. OpenAI, the parent company of ChatGPT, reaches a hefty settlement with the Department of Justice following allegations that the company and one of its subsidiaries discriminated against U.S. workers. According to the DOJ, the tech company reached a $3.2 million agreement after the agency said, preferred workers with temporary employment visas. Now, companies are typically required to undergo permanent labor certification where they advertise roles to U.S. workers. And as part of the agreement, OpenAI agreed to post open roles on its public jobs website and accept electronic applications. Of the $3 million settlement, $2 million will be for back pay fund to support victims. A new study