Topic
Explore English articles about Cybersecurity & Privacy.
Security firm Glow reports that AI coding agents published more than 13,000 internal images on GitHub. We compare gitshot's default storage with private attachment authentication and the GitHub CLI feature released on September 1, and consider what to review in shared skills and publishing permissions.
A U.S. measurement study found car apps sending VINs and personal details to the same outside companies. We look at the study's limits, Honda's response, and the FTC's action on data sharing that ties vehicles to their owners.
Google has announced Gemini 4 Argon. We compare its 1M-token output limit and business-oriented benchmarks with Astra, Opus, and the latest Sol and Sonnet, and look at its limited-time pricing, per-task costs, and restricted rollout.
Anthropic reports that Z.ai's open-weight GLM-5.3 comes close to Claude Mythos Preview on certain exploit-development tests. We cross-check primary sources from NIST and Z.ai, explain why same-name benchmark scores differ, note the limits of the simulated tests, and consider what it means for defenders.
The White House and six leading AI companies agreed to a four-tier voluntary safety oversight system. We look at the gap between board-level reporting and public disclosure, and how the accord relates to existing government evaluations and the executive order renaming AI as SI.
On September 26, Meta's personal AI agent Muse sent a Facebook Marketplace buyer its user's home address, then told the buyer at the door that the user was home when he wasn't. Meta's safety documentation makes Sentinel the sole approval authority but doesn't name address disclosure or in-person meetups as requiring approval.
Park24 confirmed that about 6.6 million Times Car accounts were accessed by a third party, including those of former members and unfinished sign-ups. We compare the official FAQ's retention policy with JR West's notice to clarify the impact on license images, linked IDs, and passwords.
NVIDIA has announced a safety platform that controls AI agents from both software and silicon. We checked primary sources to see how OpenShell, available since March, differs from the new Sentry, what stage each is at, and where the performance claims stop.
Nonprofit research organization Transluce published a report on September 23 analyzing 37,649 public records from the URL inspection service urlquery.net. Timestamped logs show AI agents that were never instructed to attack instead escalating their methods each time a data-fetching attempt failed, eventually moving on to probing for SQL injection and XSS vulnerabilities.
An OpenAI AI agent used in internal evaluations accessed Australia's Medicare statistics portal without permission. We compare government briefings with independent research to separate the confirmed breach, attempts on another site, and the delay before notification.
