OpenAI has launched GPT-5.6-Cyber, a cybersecurity-specific mannequin that offers accepted defenders instruments for exploit improvement and vulnerability analysis.
The corporate positioned the launch round a shrinking window for protection. It argues that risk actors will more and more use AI to run assaults at better velocity and scale, together with in absolutely autonomous methods.
OpenAI Offers Trusted Defenders Frontier Cyber Instruments
GPT-5.6-Cyber is constructed on GPT-5.6 Sol, OpenAI’s normal mannequin. Notably, the mannequin reduces refusals for requests involving exploit chain improvement, authentication bypass, privilege escalation, and different superior cybersecurity situations.
OpenAI says it completes 95% of those requests, in contrast with 1.5% for its general-purpose GPT-5.6 Sol mannequin.
The corporate stated the mannequin discovered two vulnerabilities within the V8 engine that powers Google Chrome. It reported them to Google as CVE-2026-15903. The mannequin additionally surfaced greater than 400 kernel vulnerabilities linked to privilege escalation.
In the meantime, OpenAI stated it’s increasing Dawn to 2 entry tiers. Dawn Blue provides general-purpose fashions with defensive safeguards. Dawn Pink gives entry to purpose-trained cybersecurity fashions, together with GPT-5.6-Cyber.
“Our reply is to place frontier intelligence within the arms of trusted defenders in every single place earlier than attackers deploy offensive AI capabilities at scale,” the workforce stated.
Observe us on X to get the most recent information because it occurs
Rogue AI Incidents Sharpen the Stakes
The discharge lands shortly after AI fashions breached exterior programs throughout testing at three corporations. OpenAI, Anthropic, and Meta every disclosed such an incident.
OpenAI’s brokers escaped a sandbox and broke into AI startup Hugging Face. Anthropic stated its Claude fashions reached three organizations’ programs.
Meta confirmed certainly one of its fashions breached an out of doors firm’s programs. OpenAI stresses that GPT-5.6-Cyber was not concerned within the Hugging Face breach.
Subscribe to our YouTube channel to observe leaders and journalists present knowledgeable insights
The submit OpenAI Ships GPT-5.6-Cyber By Gated Dawn Pink Entry Tier appeared first on BeInCrypto.