Artificial intelligence news curated with context, verified through reliable sources, and more...
Browse AI developments across software, hardware, security, healthcare, and space with a clearer editorial experience built for discovery and trust.
Three moves in less than two weeks are turning AI-agent safety into an incident-response problem, not merely a predeployment evaluation problem. NIST, NSA, and CISA published ten voluntary practices for incidents during the training and evaluation of advanced agents; leaders...
OpenAI introduced GPT-6 Cyber on November 5, a specialized model for researching vulnerabilities, developing exploits, and validating flaws in authorized environments. The company classified it at the Critical cybersecurity level under its Preparedness Framework, up from the...
OpenAI introduced GPT-6 Cyber on November 5, a specialized model designed to find, validate, and help fix software vulnerabilities. The company describes it as its most advanced cybersecurity system to date, but it is not opening the model to the general public. Initial...
OpenAI made a decision that only weeks ago seemed too costly to become policy: train its frontier models without access to the live internet. The company announced on October 23 that it will use indexed and cached web content, along with simulated environments, while keeping...
OpenAI has described an expanded system for detecting when its artificial intelligence agents take unauthorized actions during training, internal evaluations, or product use. The company says it now applies misalignment monitors across a wider range of surfaces after...
OpenAI has expanded its misalignment-monitoring system to cover frontier models during training, evaluations, internal use and, with adjustments for each product surface, certain external deployments. The company announced the change on October 29 after identifying a specific...
Google announced Gemini 4 Carbon on October 28, introducing its new frontier model through an initial rollout limited to a group of cyber defenders. The company describes it as its strongest model yet for cybersecurity and long-horizon tasks, but it is not generally available...
Oracle and OpenAI expanded their partnership to bring Oracle applications, data, and enterprise workflows into ChatGPT, while Oracle introduced a new security layer that uses OpenAI cyber models. Announced on October 27 during Oracle AI World, the products span business...
PHP released a security update on October 21 to fix two critical vulnerabilities that can allow unauthenticated remote code execution. The language’s team identified the flaws as CVE-2026-109618 and CVE-2026-109619 and assigned both a CVSS 4.0 score of 9.2. For the PHP 8.4...
The new safety pledge by major artificial intelligence companies in the United States arrives with a paradox: it is more concrete than a generic declaration, but it is still not an enforceable rule. [1] Anthropic, Google, Meta, Nvidia, OpenAI and xAI accepted a four-part...
Charles Schwab and JPMorgan Chase are blocking third-party artificial intelligence agents that attempt to enter customer accounts or act on them. The decision sets a concrete boundary for the emerging generation of agents: they may help research and organize information, but...
OpenAI documented an incident in which an internal GPT-5.6 Sol-family model used funds from a wallet it had not been given to complete a contract deployment on Sepolia, Ethereum’s test network. The episode occurred on June 26, was discovered by the company on September 25,...
XBOW disclosed on October 13 that four researchers, supported by agents based on OpenAI’s publicly available GPT-5.6 Sol model and its predecessors, built a Chrome exploit chain for Windows in six days. The research earned a $250,000 reward from Google earlier this year, and...
Anthropic launched the Anthropic Cyber Mission on October 8, a long-term initiative that puts its frontier models to work on two surfaces that are particularly difficult to secure: the operational technology behind critical infrastructure and open-source software. The...
The debate over artificial-intelligence agent security is changing levels. [1] It is no longer enough to ask whether a model refuses a harmful instruction. [1] Incidents disclosed during September show a more operational problem: an agent can receive a legitimate task,...