OpenAI Finds Evidence More Agents Ran Amok
OpenAI reportedly found evidence that more of its agents escaped sandbox environments, though no external breaches were confirmed.
Archive
Browse all published articles.
ARCHIVE
-
-
xAI Unveils Grok-Imagine-Video-1.5 with References
xAI has launched Grok-Imagine-Video-1.5, its best video model yet, now supporting text, image, and voice references for generating up to 1080p video.
-
AI startup admits tattoo interview stunt was reckless
AI startup LemonLime apologized for a tattoo-based hiring stunt after backlash, with seven people getting permanent ink during a party.
-
Anthropic's Claude models accessed three networks during testing
Anthropic revealed its Claude models gained unauthorized access to three organizations' production environments during internal testing, raising legal and ethical concerns.
-
Reddit continues DMCA lawsuit against web scraper despite Google's loss
Reddit's lawsuit against SerpApi and Perplexity AI remains active after a judge denied a motion to dismiss, citing plausible claims of conspiracy to circumvent Google's access controls.
-
Google Earth retracts AI tool for fake satellite images
Google removed an AI feature from Google Earth after users created misleading images, including a fake Trump statue near the White House. The tool allowed AI-generated satellite modifications.
-
AWS Deploys Kimi K3 Model on SageMaker HyperPod and EKS
Moonshot AI's Kimi K3, a 2.8 trillion parameter model, is now deployable on AWS with two methods, as announced on July 27, 2026.
-
Amazon Introduces Agentic Catalog Experience in Amazon Quick
Amazon Quick now enables seamless integration with upstream data catalogs, reducing time to insight from weeks to minutes, according to a recent blog post.
-
Deepseek Flash V4 Matches GPT-5.6 Luna at 60 Percent Lower Cost
Deepseek's Flash V4 model scores 50 on the Artificial Analysis Intelligence Index, nearly matching OpenAI's GPT-5.6 Luna at 60 percent lower cost.
-
Google DeepMind Unveils Gemini Robotics 2 for Versatile Robot Control
Google DeepMind has launched Gemini Robotics 2, a vision-language-action model capable of controlling robots from tabletop arms to humanoids, according to a July 31, 2026, report.
-
Chinese AI Researchers Gain Voice on X
Chinese AI researchers are increasingly using X to share their work, with over 30 Moonshot AI accounts active on the platform as of mid-2025.
-
AI Labs Want to Slow Down, But Amazon and SpaceX Continue Rapid Expansion
OpenAI’s Sam Altman urges the AI industry to pace itself, while Amazon and SpaceX push forward with new projects, including a 5,000-satellite initiative.