Solana Whale That Made $20 Million in 2023 Starts Buying Again
A Solana (SOL) whale that banked more than $20 million in 2023 has resurfaced after two years, buying $3.6 million in SOL.…
OpenAI has released GPT-5.6-Cyber, a cybersecurity-specific model that gives approved defenders tools for exploit development and vulnerability research. The company positioned the launch around a shrinking window for defense. It…
OpenAI has released GPT-5.6-Cyber, a cybersecurity-specific model that gives approved defenders tools for exploit development and vulnerability research.
The company positioned the launch around a shrinking window for defense. It argues that threat actors will increasingly use AI to run attacks at greater speed and scale, including in fully autonomous ways.
GPT-5.6-Cyber is built on GPT-5.6 Sol, OpenAI’s standard model. Notably, the model reduces refusals for requests involving exploit chain development, authentication bypass, privilege escalation, and other advanced cybersecurity scenarios.
OpenAI says it completes 95% of these requests, compared with 1.5% for its general-purpose GPT-5.6 Sol model.
The company said the model found two vulnerabilities in the V8 engine that powers Google Chrome. It reported them to Google as CVE-2026-15903. The model also surfaced more than 400 kernel vulnerabilities linked to privilege escalation.
Meanwhile, OpenAI said it’s expanding Daybreak to two access tiers. Daybreak Blue offers general-purpose models with defensive safeguards. Daybreak Red provides access to purpose-trained cybersecurity models, including GPT-5.6-Cyber.
“Our answer is to put frontier intelligence in the hands of trusted defenders everywhere before attackers deploy offensive AI capabilities at scale,” the team said.
Follow us on X to get the latest news as it happens
The release lands shortly after AI models breached outside systems during testing at three companies. OpenAI, Anthropic, and Meta each disclosed such an incident.
OpenAI’s agents escaped a sandbox and broke into AI startup Hugging Face. Anthropic said its Claude models reached three organizations’ systems.
Meta confirmed one of its models breached an outside company’s systems. OpenAI stresses that GPT-5.6-Cyber was not involved in the Hugging Face breach.
Subscribe to our YouTube channel to watch leaders and journalists provide expert insights
A Solana (SOL) whale that banked more than $20 million in 2023 has resurfaced after two years, buying $3.6 million in SOL.…
XRP slips below $1 even as a Korean regional bank adopts Ripple Payments CoinDesk
A $24 million New York Lotto ticket bought on May 25, 2016, at Renu Corp Grocery and Tobacco in Tribeca sat unclaimed…
Monad, an Ethereum rival, offered early investors up to $60 million to cash out. Almost all said no CoinDesk