AI news in context

The ChatGPT maker is moving down the silicon stack with a purpose-built ASIC it says beats current state-of-the-art on performance per watt.

• OpenAI announced Jalapeño, its first custom AI inference processor, developed in partnership with Broadcom and Celestica.
• The ASIC is purpose-built for LLM inference, with early testing claiming performance per watt substantially better than current state-of-the-art.

A fresh CNET review and hands-on reports confirm NotebookLM remains a uniquely practical AI tool after its largest update to date.

• CNET published a positive review of Google's NotebookLM, calling it a uniquely practical AI tool for working with your own sources.
• Google recently shipped what it calls NotebookLM's biggest update yet, adding visible reasoning so users can see the model's intermediate steps.

The India-based marketing platform's all-cash deal points toward a future where every customer gets their own AI agent.

• MoEngage completed an all-cash acquisition of technology that assigns individual AI agents to individual customers.
• The deal signals a strategic bet that marketing's future is agent-per-customer personalization at massive scale.

Netflix, A24, Focus Features, and Warner Bros.' Clockwork have all declined to distribute Luca Guadagnino's nearly completed film, leaving Mubi and Neon as the remaining suitors for a project caught between tech and entertainment.

• Amazon MGM Studios abruptly dropped the nearly completed film after chief Mike Hopkins screened a cut; no substantive public explanation has been given.
• Amazon's exit came roughly four months after announcing a $50 billion strategic partnership with OpenAI on February 27, 2026.

Backed by Hollywood A-listers and launched at the European Parliament, the registry lets anyone set machine-readable terms for how AI systems use their name, image, voice, and likeness.

• RSL Media, co-founded by Cate Blanchett, launched the free Human Consent Registry on June 23, 2026, at the European Parliament.
• The registry lets anyone declare whether AI systems may permit, prohibit, or pay for use of their name, image, voice, and likeness.

The productivity platform formerly known as Grammarly is snapping up the Princeton-born AI detector with 19 million users and $30M in ARR.

• Superhuman (formerly Grammarly) is acquiring GPTZero, an AI content detection startup with 19 million users and $30M ARR; deal terms were not disclosed.
• GPTZero was valued at over $88M per PitchBook; backers include Uncork Capital, Neo, Footwork, and Jack Altman.

Two rival AI-industry super PACs are spending heavily in New York's 12th District, turning a local Democratic primary into the first congressional battleground over AI regulation.

• Over $20 million in AI-industry PAC spending has poured into New York's 12th District Democratic primary — the only congressional race where both major AI super PACs are involved.
• Leading the Future (backed by Andreessen Horowitz, Greg Brockman, Perplexity) spent $8M opposing AI safety advocate Alex Bores; Public First Action (backed by $20M from Anthropic) is supporting him.

The new model doubles the clip-length norm, accepts 50 references, and lets you edit specific elements without regenerating the whole scene.

• ByteDance unveiled Seedance 2.5 at Volcano Engine's FORCE conference, claiming 30-second single-prompt video generation without stitching shorter clips.
• The model supports native 4K output and up to 50 image, video, and audio references per request.

Anthropic has launched Claude Tag in beta for Enterprise and Team customers, replacing its prior Slack app with a persistent AI agent that learns, monitors, and works autonomously inside channels.

• Anthropic launched Claude Tag in beta for Claude Enterprise and Team customers, replacing its prior Claude in Slack app.
• Users can tag @Claude in channels and threads to delegate tasks; the AI functions as a persistent, shared teammate with memory.