News
Updates, product notes and practical coverage for AI platforms and operators.
OpenAI is partnering with Hacktron to test high-priority systems and third-party integrations, while expanding the security firm’s access to frontier models through Daybreak Red. The arrangement combines an established bug-bounty relationship with targeted collaborative research and forms part of a broader push to find and fix vulnerabilities before attackers can exploit them.
OpenAI says an autonomous Linux-kernel security pipeline reproduced 1,724 vulnerabilities and developed 948 distinct local privilege-escalation exploits with candidate patches. The project spent about US$1.21 million in API credits and shows how agentic security research can scale, while also exposing the need for aggressive deduplication, reproducibility and expert patch review.
OpenAI has made GPT-5.6 Cyber available to approved Daybreak Red customers through Amazon Bedrock in AWS GovCloud US-East. The controlled service gives eligible government security teams stronger support for vulnerability research, exploit validation and mitigation work while retaining AWS provisioning, identity checks, monitoring and explicit human review requirements.
OpenAI has introduced GPT-6 Cyber, a restricted cybersecurity model for vulnerability discovery, exploit validation and patch development. It improves ExploitGym performance by 51% relative to GPT-5.6 Cyber and has already supported Firefox and Linux research. Initial access is limited to approved Daybreak Red customers through the API and Codex.
Mistral has released open weights for Mistral Large 4, its one-trillion-parameter multimodal mixture-of-experts model. The updated release adds three reasoning modes and reports stronger coding, cybersecurity and token-efficiency results. Weights are available for self-hosting under a modified MIT licence, while API pricing starts at US$0.70 per million input tokens.