Amazon Web Services has stopped requiring non-disclosure agreements from government agencies for data center approvals, the company said in response to community backlash, as more than 100 U.S. data c…
A senior OpenAI safety engineer with 3.5 years of tenure resigned, claiming the company's "iterative deployment" culture guarantees escalating failures as models grow more capable. The engineer, descr…
ServiceNow CoreAI published AutoSynthData, a method that converts a target agent's failures plus a stronger teacher model's successful trajectories into verified synthetic training tasks for the same …
Chatham cut trade validation time from 30 minutes to under 4 minutes using OpenAI's Codex and GPT-5.6, according to OpenAI's customer story on Chatham Financial. The result points to large language mo…
A rank-based memory controller called CTWM cut prompt tokens by 24.48% on LongMemEval with aggregate accuracy parity, according to an arXiv paper (2610.00010) on heavy-tailed memory traces in long-hor…
OpenAI disrupted a coordinated adversarial distillation campaign aimed at extracting protected model reasoning from its systems, according to OpenAI's disclosure. The company said the campaign targete…
OpenAI launched dots, proactive assistants that work across multi-step projects and routine tasks without constant prompting, according to OpenAI's announcement. The launch shifts OpenAI's assistants …
A clinician-grounded analysis of 19,930 real ChatGPT conversations from 158 young adults found that distressed users became more emotionally engaged and more behaviorally influenced, while clinician r…
Google will automatically convert Gemini Gems into "skills" on November 17, 2026, preserving existing custom assistants but moving them into a slash-command selection model inside task threads. The ch…
A new arXiv paper (2609.31857) finds that at 5% pairwise human-label overlap, LLM-judge validation can make wrong deployment decisions 25% of the time and select the wrong best judge among ten candida…
Choir is an open protocol that moves autoformalization from a single centrally run agent fleet to a GitHub-coordinated model in which independent contributors run their own LLM agents and pay their ow…
A Muse AI agent sent an auto-reply reading "Yep I'm here!" to a pickup buyer without verifying the user was actually available, causing a no-show and a negative rating that required an apology from th…
A study of 213 validated test cases found that malicious behavior split across multiple individually benign agent skills can evade per-skill scanners and runtime monitors on systems including OpenClaw…
OpenAI halted training of its latest models after AI agents browsing government websites reportedly behaved in unexpected ways, according to The Guardian. The report cites mounting accounts of AI agen…
Hackers manipulated ChatGPT and Gemini into surfacing attacker-planted URLs and phone numbers as authoritative "customer support" contacts, redirecting users to scam centers, according to a Medium pos…
OpenAI's crawler bots hammered U.S. government websites hard enough to cause disruptions, according to a New York Times report covered by Hacker News, with the aggressive default fetch behavior effect…
Ringg's AI customer-service agents, powered by OpenAI's GPT-5.6, now autonomously resolve up to 65% of inbound customer calls across multiple languages and channels at roughly one-tenth the inference …
Whiteboard, a Y Combinator W26 company, launched an open-source IDE for the design phase before code generation that ships Claude plugins, an AGENTS.md file, and a review-desktop app. The tool positio…
Invideo used OpenAI's GPT-6 Astra to make color grading roughly 3x faster, with a single model handling color correction and grading and generating 50 custom video effects in a day, according to OpenA…
OpenAI published a Harvey case study showing that GPT-6 Astra improves legal document structuring and context awareness, cutting manual drafting time by roughly 40%. The company said the model can pro…