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·Apr 14·5 min read
Trusted access for the next era of cyber defense
TL;DR
Cybersecurity defenders previously lacked access to advanced AI models capable of identifying sophisticated threats, creating a gap between offensive and defensive capabilities. OpenAI introduced GPT-5.4-Cyber, a specialized model deployed through its Trusted Access for Cyber program to vetted security professionals.
✦ Why It Matters
Engineers can now access GPT-5.4-Cyber through proper channels to accelerate threat detection and incident response workflows.
Key Takeaways
- Request access to GPT-5.4-Cyber through OpenAI's vetting process if your organization defends critical infrastructure.
- Integrate the model into incident response workflows to accelerate threat analysis and vulnerability assessment.
- Design safeguards into your deployment to prevent unauthorized use or misuse of AI capabilities.
- Monitor model outputs for accuracy and bias in your specific threat environment before full deployment.
Full Summary
As artificial intelligence advances, cybersecurity threats grow more sophisticated, but defenders have struggled to access equally powerful AI tools for threat detection and response. OpenAI addressed this gap by expanding its Trusted Access for Cyber program, which provides vetted cybersecurity professionals with controlled access to advanced AI capabilities.
The program introduces GPT-5.4-Cyber, a specialized large language model (LLM) fine-tuned for cybersecurity tasks like vulnerability analysis, threat intelligence synthesis, and incident response planning. Access is restricted to vetted defenders through a vetting process that ensures responsible use.
The expansion includes strengthened safeguards—technical and policy-based controls—designed to prevent misuse while enabling legitimate defensive work. This approach balances enabling innovation in cyber defense with mitigating risks of AI-enabled attacks.