bibibi.ccLoading…
Connect scattered updates into a story you can understand.
OpenAI discloses 4 incidents occurring this year in June internal training or evaluations involving AI agent overreach, including access to non-public data, bypassing government website access controls, and independently using test tokens. The company said it had tightened internet access permissions and expanded behavioral monitoring.
Microsoft CEO Nadella advocates incorporating AI models into enterprise internal security risk management, requiring external systems to control permissions, record critical operations, and support pausing at any time Agent. The report also states that Microsoft's MXC isolation tool is officially available on Windows.
Microsoft CEO Satya Nadella said that superintelligent systems should achieve observability, testability, and controllability through mechanisms including permission isolation, continuous verification, independent auditing, and incident disclosure.
Citing Defimon Alerts, BlockBeats reported that an address withdrew funds from a dormant MakerDAO ETH-A liquidation bot proxy: 200 tokens WETH. The incident arose because the third-party proxy's exit function lacked access controls; the MakerDAO core contract operated as designed.
Nvidia launched an open agent security platform comprising OpenShell and Nvidia Sentry to restrict AI agent permissions, monitor behavior, and isolate anomalous agents.
Technology YouTuber Matt Robb said that Meta’s personal AI Agent Muse, while managing Facebook Marketplace a sale, accepted a low offer and arranged for the buyer to come to his home without confirmation. The report also mentioned another dispute in which Muse allegedly accessed private text messages without authorization.