We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Oct 7·5 min read
Disrupting malicious uses of AI: October 2025
TL;DR
AI systems face misuse risks including malicious applications that cause real-world harm, but detection and enforcement mechanisms remain incomplete. OpenAI deployed detection systems and policy enforcement tools to identify and disrupt harmful uses of their models across multiple threat categories.
✦ Why It Matters
Engineers can adopt detection patterns and enforcement workflows to protect their own AI systems from coordinated misuse.
Key Takeaways
- Implement automated detection for API usage anomalies and behavioral patterns indicative of misuse.
- Combine automated flagging with human review to reduce false positives in enforcement decisions.
- Document and measure enforcement actions to track effectiveness and identify emerging threat patterns.
- Coordinate with external parties and law enforcement when malicious activity poses systemic risks.
Full Summary
AI platforms must balance accessibility with safety, as large language models and generative systems can be repurposed for harmful activities including fraud, deception, and illegal operations. OpenAI implemented detection systems that monitor API usage patterns, user behavior, and output characteristics to identify misuse in real time.
The approach combines automated flagging with human review, policy enforcement (account suspension, API restrictions), and coordination with external parties. The October 2025 report details specific disruption cases across threat categories such as scams, unauthorized access attempts, and coordinated inauthentic behavior.
Results include quantified enforcement actions and documented cases where malicious campaigns were interrupted before scaling. These findings inform how AI providers can operationalize safety at scale while maintaining transparency about enforcement decisions and their limitations.