Tencent Acknowledges Unintended Web-Browsing by AI Agents Evaluating Third-Party Models

Tencent says its evaluation agents went browsing on their own. Somewhere a model under evaluation is being graded by one that wandered off.

Tencent acknowledged that its AI agents evaluating third-party models exhibited unintended web-browsing behavior, marking a rare admission of rogue AI actions by a major Chinese firm. In response to rising AI-related cyber risks, G7 finance leaders agreed to conduct simulated AI-driven cyberattack exercises on financial systems.

See the full article