Unbiased AI-powered news
OpenAI announced Daybreak, an initiative that uses AI to detect and patch software vulnerabilities before they can be exploited. The system relies on the Codex Security AI agent launched in March along with specialized models including GPT-5.5-Cyber. The launch follows a rival company's announcement last month of a similar but more restricted security-focused project.
The VergeOpenAI announced the launch of Daybreak, an AI initiative designed to detect and patch vulnerabilities in software before attackers can find them. The system relies on the Codex Security AI agent, which the company introduced in March. That agent creates a threat model based on an organization's code, identifies possible attack paths, validates likely vulnerabilities and automates detection of the higher-risk ones.
Daybreak combines multiple OpenAI systems. It brings together the company's most capable models, the Codex agent and security partners. It also incorporates specialized cyber models including GPT-5.5 with Trusted Access for Cyber and GPT-5.5-Cyber, which began rolling out last week.
The announcement arrives shortly after another company introduced a security-focused AI model last month. That model was described as too dangerous for public release and was shared only privately as part of a restricted initiative. Some unauthorized parties later gained access to it.
OpenAI had previously lacked a comparable dedicated security product. The company stated that Daybreak brings together the most capable OpenAI models, Codex, and its security partners.
OpenAI reported that it is working with industry and government partners. The company said it is preparing to deploy increasingly more cyber-capable models in the future. The initiative represents the company's entry into a specialized area of AI application focused on proactive cybersecurity measures.
Single source — no framing comparison available.
sbs.com.auA Meta Oversight Board study released Thursday found that 10 major large language models were more likely to refuse requests to criticize leaders in countries with strict speech laws. The results applied to prompts about leaders in China, Saudi Arabia, Thailand and Turkey.
Nvidia introduced Cosmos 3 Edge, a world model for real-time physical environment navigation. The launch occurred as CEO Jensen Huang visited Japan to form industrial partnerships.
wccftech.comNoetra will oversee the project with ¥387.3 billion in funding and build a 140-megawatt data center. The effort draws engineers from SoftBank, NEC and other firms to develop a domestic AI system for robotics.