Live Updates > Live Update Details
2026-11-04 21:14:26
[Israeli Security Company Says Some AI Cyber Intrusions Were Due to Human Error] At a recent tech industry event in London, Dan Rahaf, CEO of the Israeli AI company Unconventional, said that during a series of security tests on AI models conducted by his company, human error and time pressure led to some isolation and protection measures being improperly configured, allowing the models to access the internet and commit cyber intrusions. According to Bloomberg, in the past six months, US companies Anthropic, OpenAI, Metaverse, and Google have all disclosed cyber intrusion incidents involving model security testing, which was overseen by Unconventional. Such assessments aim to help developers identify risks and develop protective measures before releasing models to the public. Rahaf reportedly stated at the Bloomberg Tech London conference on the 3rd that the issues the company encountered earlier this year were resolved on the day of the incident. (Xinhua)
