As AI-led attacks multiply, OpenAI launches a new cyber model
THE SO WHAT
Defensive AI is professionalizing—OpenAI’s new cyber-trained model under Daybreak means some blue teams will get access to offense-grade capabilities with guardrails. CISOs should treat this as a new category in the stack and start piloting with clear policies on scope, logging, and human review.
READ THE SOURCE
MORE FROM THE WIRE
Applied AITencent’s WorkBuddy Sparks AI Turnaround Hopes After Stock Rout
If WorkBuddy becomes Tencent’s wedge back into enterprise AI, it reinforces a pattern: productivity copilots as the recovery vector for consumer-heavy platforms. For operators in China, assume your internal tooling will be compared directly to WorkBuddy’s UX and integration depth over the next 6–12 months.
Applied AIAnthropic Strikes $9 Billion Cloud Deal With Riot Platforms
A $9.1 billion deal with a former Bitcoin miner turned AI data center vendor shows how quickly stranded energy and power-heavy infra are being re-rated as model capacity. If you’re compute-constrained, it’s time to map nontraditional partners—miners, industrial parks, energy developers—not just hyperscalers.
Applied AIOpenAI launches GPT-5.6-Cyber with reduced refusals, 95% completion on advanced cybersecurity tasks
A model explicitly tuned for vulnerability research and exploit development—at 95% completion rates and fewer refusals—moves serious offensive capability into a product SKU. Security leaders need to assume both defenders and adversaries will have similar tools and shift focus to detection speed, containment, and model access governance.
Applied AISource: OpenAI head of ethics Chloé Bakalar departs after less than a year, following exits by the head of safety systems and a former head of mission alignment
Multiple senior departures in ethics, safety systems, and mission alignment at a major lab will sharpen external scrutiny on how safety is operationalized versus staffed. Enterprise buyers should respond with specifics: ask for current org charts, escalation paths, and how safety decisions map into product roadmaps and SLAs.