Model Security

Model security encompasses the technical and operational measures that protect machine learning systems from unauthorized access, tampering, and malicious attacks. Organizations must address vulnerabilities across the model lifecycle, including training data poisoning, inference-time adversarial attacks, and model extraction threats that could compromise proprietary AI systems or enable harmful outputs. For enterprise AI governance, robust model security controls are essential to prevent intellectual property theft, ensure regulatory compliance, and maintain stakeholder trust in AI-driven decision-making systems.

1 item

ResearchGlobal2026-08-11

Frontier API Reasoning Traces Leaked 62 Live API Keys in Public Agent Logs

Researchers from MATS Research, the ELLIS Institute Tubingen, and the Max Planck Institute for Intelligent Systems published findings showing that encrypted chain-of-thought reasoning blocks returned by Anthropic, OpenAI, and Google APIs can be replayed across sessions and users to extract hidden plaintext reasoning. Analysis of 6,708 publicly available agent trajectory logs reconstructed 315,320 reasoning blocks containing 704 distinct privacy artifacts, including 62 API keys, 33 passwords, and 24 access tokens. Sixty-four of those artifacts appeared exclusively inside reasoning blocks invisible to end users, meaning standard output-layer DLP controls would not have detected them.

Claude Opus Anthropic OpenAI Google DeepMind API risk credential leakage chain-of-thought model security