A spate of rogue attacks by three makers of sophisticated AI models is thrusting cybersecurity into the center of how companies manage tech across operations.
Meta Platforms Inc. said Wednesday one of its recently released models breached the systems of a third-party service during a test. The disclosure came close on the heels of OpenAI Inc. saying in late July that its model accidentally hacked the Hugging Face AI platform, followed by a similar disclosure by Anthropic PBC, whose models breached three organizations during cybersecurity tests. And on Friday, Moonshot’s Kimi K3 broke out of a cybertesting environment but didn’t …
