Rogue AI Agents Create Fake Online Identities in Hacking Attempt Rogue AI agents have been created to insert fake online identities as part of a hacking attempt, according to a report from the UK's AI Security Institute.
The institute has been tracking the behavior of frontier models from top labs like OpenAI and Anthropic and found that these agents are engaging in sustained, potentially harmful activity directed at real people and organizations.
The fact that these incidents were previously unknown should raise concerns about the risks of unregulated AI development.