Jump within the current audio brief. Story thumbnails mirror the cards below when available.
Same-day briefs for Wed Jul 8, sorted by release time. The selected date stays fixed.
Changing the date preserves the current brief type: AI News PM.
/ai-news-pm/2026-07-08.html.Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
OpenAI rolled out GPT-Live on Wednesday, replacing turn-by-turn voice mode with a full-duplex architecture that listens and speaks simultaneously. Two versions are live globally — GPT-Live-1 for paid ChatGPT subscribers (Go, Plus, Pro) and GPT-Live-1 mini for free users — with complex queries delegated to GPT-5.5 in the background.
SpaceXAI released Grok 4.5 on Wednesday, built on its V9 foundation model at 1.5 trillion parameters and priced at $2 per million input tokens, $6 per million output. Musk claims twice the token efficiency of comparable models and Opus-level performance; public rollout begins July 9, with EU availability expected mid-July.
The European Systemic Risk Board issued a formal warning Tuesday that frontier AI models can now autonomously discover vulnerabilities and execute full cyberattacks at speed and scale beyond any prior AI system. The ECB told major EU banks to submit concrete cyber-defense action plans to supervisors by October 31, 2026.
Anthropic released a self-hosted containerized gateway that lets enterprises run Claude Code on Amazon Bedrock or Google Cloud with centralized SSO (via Okta, Entra ID, or Google Workspace), per-user spend limits, and multi-provider failover. Inference data stays on-premises and never reaches Anthropic unless explicitly configured otherwise.
NVIDIA released Nemotron-Labs-Audex-30B-A3B, a 30-billion-parameter MoE model unifying speech recognition, translation, text-to-speech, and audio generation while preserving the text reasoning of its Mamba-Transformer backbone. Both the 30B and a smaller 2B variant are on Hugging Face under a noncommercial research license.
Z.ai's 753-billion-parameter GLM-5.2 recorded 80x customer growth and 27x daily token volume on Vercel's AI Gateway in a single week, scoring 62.1 on SWE-bench Pro versus GPT-5.5's 58.6 — at roughly one-sixth the cost. The numbers underscore how decisively Chinese open-weight models are capturing developer workflow share.
Anthropic's updated privacy policy took effect Wednesday, allowing the company to require government-issued photo ID and facial biometric scans from some claude.ai Free, Pro, and Max users via identity provider Persona. The requirement flows from export-control pressure: Commerce Secretary Howard Lutnick flagged concern over adversarial military access to US frontier AI from abroad. Enterprise and API customers are exempt.
The startup, backed by General Catalyst, Jeff Bezos, and Eric Schmidt, is building physical AI foundation models trained on millions of hours of video game controller input rather than costly real-world robot experiments. Its model already adapts to a quadrupedal robot after just eight minutes of real hardware fine-tuning. TechCrunch published a deep-dive Wednesday framing it as robotics' ChatGPT moment.
