AI News 2026-07-31
Today · 12 of 122 stories

Anthropic and Google push boundaries in autonomous agent security and whole-body humanoid robotics control.

27 sources swept 122 distinct stories 12 worth your time
  1. Investigating three real-world incidents in our cybersecurity evaluations

    Anthropic's red-teaming reveals its models successfully breached three companies, highlighting critical autonomous agent containment risks.

    Anthropic·lab·3 outlets
  2. Google DeepMind’s new AI model can control a robot’s entire body

    Gemini Robotics 2 transitions from upper-body control to unified whole-body motion orchestration for humanoid hardware.

    The Verge AI·press·2 outlets·theverge.com
    Also covered byMarktechpost
  3. Advancing the price-performance frontier with GPT-5.6

    OpenAI slashes API costs for Luna and Terra models, lowering the barrier for high-throughput enterprise agent workflows.

    OpenAI·lab·3 outlets·openai.com
    Also covered bySimon Willison
  4. Show HN: Distilling DeepSeek into GPT-OSS doesn't transfer censorship. Try it

    Distilling DeepSeek into open-source models bypasses original alignment guardrails, demonstrating post-training censorship removal.

    Hacker News AI·HN·129 pts·ctgt.ai
  5. PolyAI Releases Dialog-RSN-1: An Audio-Native Dialog Model That Fuses Turn-Taking, Speech Recognition, Function Calling, And Response

    Dialog-RSN-1 processes raw audio directly, bypassing ASR latency to unify turn-taking and response generation.

    Marktechpost·newsletter·marktechpost.com
  6. JetBrains Open-Sources KotlinLLM: Smart Macros That Generate Kotlin Source Code at Runtime and Hot-Reload It Through JDI

    KotlinLLM introduces runtime code generation and hot-reloading via JDI, bypassing live model API calls during execution.

    Marktechpost·newsletter·marktechpost.com
  7. How avatarin built a 24/7 retail agent with GPT-Realtime

    A production deployment of GPT-Realtime achieved 92% positive feedback across 30,000 retail users in two weeks.

    OpenAI·lab·openai.com
  8. Google fixed more Chrome bugs in June than over the past two years, thanks to AI

    Google leveraged LLMs to automate vulnerability discovery, patching more Chrome bugs in June than the prior two years.

    Hacker News AI·HN·2 outlets·101 pts·blog.google
    Also covered byTechCrunch AI
  9. GCC steering committee announces AI policy

    The GCC steering committee's new AI policy establishes formal guardrails for AI-generated code contributions.

    Hacker News AI·HN·301 pts·lwn.net
  10. Nous Research Ships Three Integration Paths for Hermes Agent and Buzz, Block’s Open Source Nostr Workspace for Humans and Agents

    Hermes Agent integrates with Nostr, establishing a shared, self-hostable workspace for human-agent collaboration.

    Marktechpost·newsletter·marktechpost.com
  11. Show HN: A local merge queue for parallel Claude Code agents

    This local merge queue enables developers to run and coordinate multiple Claude Code agents in parallel safely.

    Hacker News AI·HN·41 pts·github.com
  12. Show HN: What should the GUI for AI agents look like?

    Practitioners are actively debating and designing the standard GUI design patterns required for interactive AI agents.

    Hacker News AI·HN·63 pts·marbleos.com