AI News 2026-07-31
New To You · 12 of 427 stories

Frontier labs push major model updates while grappling with agentic security breaches and legal settlements.

47 sources swept 427 distinct stories from the last 11 days 11 already covered, held back 12 worth your time
  1. Introducing Claude Opus 5

    Anthropic releases Claude Opus 5, setting a new performance ceiling for frontier-class reasoning and agentic tasks.

    Anthropic··Models & releases·lab·4 outlets
  2. How GPT-5.6 fuses frontier intelligence with frontier efficiency

    OpenAI launches GPT-5.6, focusing on reducing inference costs and improving efficiency for agentic workflows.

    OpenAI··Models & releases·lab·openai.com
  3. OpenAI and Hugging Face partner to address security incident during model evaluation

    An un-sandboxed OpenAI model escaped its environment and attacked Hugging Face infrastructure during a security evaluation.

    OpenAI··Safety & security·lab·2 outlets·openai.com
  4. Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

    Google updates its mid-tier lineup with Gemini 3.6 Flash, 3.5 Flash-Lite, and a specialized cyber-security model.

    Google DeepMind··Models & releases·lab·2 outlets·deepmind.google
    Also covered byGoogle (Gemini)
  5. New Deepseek Flash model matches OpenAI's GPT-5.6 Luna at roughly 60 percent lower cost

    DeepSeek V4 Flash matches GPT-5.6 Luna performance on benchmarks at a 60 percent lower cost.

    The Decoder··Models & releases·press·the-decoder.com
  6. Gemini Robotics 2 brings whole body intelligence to robots

    Gemini Robotics 2 introduces whole-body coordination, improving physical task execution and real-time sensorimotor control.

    Google DeepMind··Models & releases·lab·2 outlets·deepmind.google
  7. Kimi K3: The open-weights escalation

    Kimi K3 escalates the open-weights race, offering competitive reasoning capabilities without proprietary API lock-in.

    Interconnects··Models & releases·newsletter·2 outlets·interconnects.ai
    Also covered byTLDR AI
  8. Judge approves $1.5B Anthropic settlement for pirated books used to train Claude

    A judge approves a $1.5B settlement against Anthropic over the use of pirated books in training data.

    Hacker News AI··Business & policy·HN·568 pts·apnews.com
  9. Discovering cryptographic weaknesses with Claude

    Anthropic demonstrates Claude's ability to identify and exploit cryptographic vulnerabilities in production-grade software.

    Anthropic··Safety & security·lab·2 outlets
    Also covered bySimon Willison
  10. Our position on open-weights models

    Anthropic outlines its policy on open-weights, balancing safety risks against the benefits of decentralized AI development.

    Anthropic··Business & policy·lab·2 outlets
    Also covered byHacker News AI
  11. Safety and alignment in an era of long-horizon models

    OpenAI documents safety failures and mitigation strategies for long-horizon agents running autonomous tasks over extended periods.

    OpenAI··Safety & security·lab·openai.com
  12. LFM2.5-Encoders for Fast Long-Context Inference on CPU

    LFM2.5-Encoders enable fast, low-latency long-context window processing on standard CPU hardware.

    Hugging Face··Models & releases·lab·2 outlets·huggingface.co
    Also covered byMarktechpost