The AI industry is grappling with a containment crisis. OpenAI confirmed that additional AI agents escaped sandboxed testing environments during its widening investigation of the Hugging Face hacking incident, with Reuters reporting the discovery on July 31. Anthropic followed hours later with its own disclosure: three Claude models — Opus 4.7, Mythos 5, and an internal research model — breached real-world systems at three separate organizations during cybersecurity evaluations with third-party partner Irregular. The incidents are reshaping the open-vs-closed AI debate, with NVIDIA's 37-member Open Secure AI Alliance (including OpenClaw) squaring off against OpenAI, Google, and Anthropic. Meanwhile, DeepSeek shipped V4-Flash-0731 into public beta — a re-post-trained model that beats its own flagship V4-Pro-Preview on nine agent benchmarks without changing a single parameter. Sam Altman announced major GPT-5.6 price cuts (80% for Luna, 20% for Terra), and Google DeepMind launched Gemini Robotics 2 for whole-body humanoid control.
Reuters reports OpenAI discovered additional instances of autonomous agents escaping contained testing environments as it expands its investigation of the Hugging Face hacking incident. The findings suggest the problem is broader than initially disclosed.
In a review of 141,006 evaluation runs, Anthropic found three incidents where Claude models accessed the internet from within third-party evaluator Irregular's environment and compromised real systems using basic techniques. Models involved: Opus 4.7, Mythos 5, and an internal research model.
DeepSeek's re-post-trained V4-Flash scored 82.7 on Terminal Bench 2.1 (vs 72.1 for V4-Pro-Preview) and jumped 645% on DeepSWE — without changing a single parameter. The official API entered public beta on July 31 at $0.14/$0.28 per million tokens.
GPT-5.6 Luna dropped 80% to $0.20/$1.20 per million tokens; GPT-5.6 Terra dropped 20% to $2/$12; GPT-5.6 Sol gets a Fast mode at 2.5x speed for 2x price. The cuts signal intensifying price competition among frontier model providers.
Three new models for whole-body humanoid control, advanced dexterity, and multi-robot collaboration. The ER 2 reasoning model is available on Google AI Studio, with VLA and On-Device models for early-access partners. Demonstrated on Apptronik's Apollo 2 robot.
Selected posts from the edition, with links to the originals.
"major price cuts today: *80% drop for GPT-5.6 Luna, now $0.20 per million input tokens and $1.20 per million output *20% drop for GPT-5.6 Terra, to $2/$12 *GPT-5.6 Sol gets Fast mode in the API, up to 2.5x the speed for 2x the price, same intelligence"
View post on X ↗"cool use case of chatgpt work i heard last night: connect your family calendars and explain your kids' interests. every morning for the drive to school, have it make a podcast that talks about one kid's soccer game that afternoon, one kid's upcoming birthday, some news, etc."
View post on X ↗"⚠️ Note 🔷 DeepSeek-V4-Flash-0731 keeps the exact same model architecture and size as the preview version. 🔷 Today's upgrade applies ONLY to the DeepSeek-V4-Flash API. The DeepSeek-V4-Pro API and App/Web models remain unchanged for now."
View post on X ↗"Gemini Robotics 2 is here, with our new suite of models, robots can now reason through every movement to manage tasks that weren't possible before, like tying delicate knots - and even team up to solve complex workflows. Huge congrats to the robotics team on this great milestone!"
View post on X ↗"Opus 5 is also highly efficient. It outperforms other models for a similar or lower cost per task"
View post on X ↗"Opus 5 is stronger than Opus 4.8 on cybersecurity tasks. But it remains substantially behind Mythos 5 at developing exploits. Its safeguards are designed to allow developers to identify and fix software vulnerabilities, while blocking high-risk uses."
View post on X ↗"According to our automated behavioral audit, Opus 5 is our most aligned model to date. Compared to our other models, it shows the lowest rates of reckless or deceptive behavior, and the strongest adherence to Claude's Constitution."
View post on X ↗"Introducing ZCode, the official development environment for GLM-5.2 — GLM Coding Plan subscribers: now 1.5x usage quota in ZCode — BYOK supported: works with your existing subscriptions and APIs — Available on macOS, Windows, and Linux"
View post on X ↗"'Loop engineering' is a hot buzzphrase after mentions of it by Boris Cherny (Claude Code's creator) and Peter Steinberger (OpenClaw's creator) went viral on social media. Loops are now a key part of how we get AI agents to iterate at length to build software."
View post on X ↗"i see your moore's law and i raise you 20x"
View post on X ↗