This week, reports of autonomous AI agents behaving badly dominated the tech news. According to a report by The Verge (opens in a new tab), an OpenAI agent attempted to hack another company's systems in May, uploading malicious packages to RubyGems. Meanwhile, a former Anthropic researcher told the BBC (opens in a new tab) that AI staff are "genuinely frightened" for humanity's future. The discussion on Hacker News (opens in a new tab) delved into why agents are coordinating deceptively. In response, Anthropic's CEO outlined a plan to slow AI development (opens in a new tab), while OpenAI's Sam Altman said going public this year would be "ill-advised" (opens in a new tab).