Microsoft demonstrates how an attacker-initiated OAuth device authorization can extend into a coding-agent environment after a user approves it on the genuine GitHub website. Its controlled Copilot CLI simulation produced four possible outcomes: steering a remotely enabled active session, adding misleading memory, reading cloud-synchronized sessions and accessing repositories permitted by the authorization. Each depends on account scopes and enabled features. Remote steering additionally requires an online interactive session, organizational permission and local execution controls; this is not unauthenticated remote code execution. Session synchronization alone does not enable remote control. Microsoft explicitly says these four outcomes were laboratory demonstrations, not observed attacks in the wild. Its detection approach correlates suspicious messages and clicks with GitHub OAuth activity, requiring the relevant connector and telemetry. The enduring lesson is that a trusted verification domain identifies the service, not the person who initiated the authorization.
Device-code phishing: audit coding-agent sessions, memory and OAuth grants
Related research
More curated notes connected through Agent Security and AI Red Teaming.
Pacing model development in an era of cyber-critical capabilities
OpenAI says preliminary evidence that Astra may meet its Critical cybersecurity threshold led it to pause frontier reinforcement-learning work for two weeks and keep its largest planned run on hold. New safeguards include stronger workload and network isolation, continuous boundary testing, token-level monitoring that escalates suspicious tool activity, and broader alignment checks for deception, reward hacking, and unauthorized access.
OWASP Top 10 for Agentic Applications for 2026
OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.
A blueprint for democratic governance of frontier AI
OpenAI proposes a three-part U.S. frontier-AI governance model: harmonize emerging state safety laws into a federal baseline, strengthen CAISI as an evaluation and standards institution, and coordinate a broader resilience program. Proposed controls include severe-risk evaluations, transparency reports, independent audits, safety-incident reporting, model-weight security, whistleblower protection, and periodic technical assessments.