Tag
The OpenAI-Hugging Face incident is highlighted as a pivotal moment in cybersecurity, showing how AI models are evolving to automate both attacks and defenses, with a call for organizations to urgently upgrade their security practices using AI.
OpenAI is expanding its Daybreak cyber defense service with two tiers, Blue and Red, and releasing a new defensive cyber model, GPT-5.6-Cyber, available only to trusted partners as AI-driven attacks rise.
Tracebit introduces 'context bombing,' a technique that uses prompt injections as defensive decoys to halt AI hacking agents, reducing admin compromise from 57% to 5% across tests with leading LLMs.
OpenAI launches Trusted Access for Cyber, an identity and trust-based framework piloting access to GPT-5.3-Codex for defensive cybersecurity work, alongside a $10 million API credit commitment to accelerate cyber defense capabilities while mitigating misuse risks.