AI Morning Brief
Top AI industry developments, filtered for what matters in mobility and automation.
Anthropic Withholds Claude Mythos — Deemed Too Powerful for Public Release
Anthropic announced that its newest frontier model, Claude Mythos Preview, will not be made generally available due to its exceptional ability to find high-severity cybersecurity vulnerabilities in major operating systems and web browsers. Instead, 12 partner organizations including AWS, Apple, Google, Microsoft, and NVIDIA will get access through 'Project Glasswing,' a defensive cybersecurity initiative backed by $100M in usage credits.
Why it matters: This marks the first time a major AI lab has restricted a model's release purely for cybersecurity reasons. The move signals a new era where AI capabilities in offensive security outpace defensive readiness — and sets a precedent for how frontier labs may handle increasingly capable models.
Meta Unveils Muse Spark — First Model from Superintelligence Lab
Meta released Muse Spark, its first AI model from the newly formed Superintelligence Lab led by former Scale AI CEO Alexandr Wang. The model features advanced reasoning, multi-agent 'Contemplating' mode, health expertise trained with 1,000+ physicians, and a unique shopping mode. The Meta AI app surged to No. 5 on the App Store following launch.
Why it matters: Meta's $14.3B investment in Scale AI and recruitment of Wang signals a serious pivot. Muse Spark's multi-agent architecture and integration across Facebook, Instagram, WhatsApp, and Ray-Ban glasses could bring advanced AI to Meta's billions of users — a distribution advantage no competitor can match.
OpenAI Launches $100/Month Pro Plan, Halving Previous Price
OpenAI introduced a new $100/month ChatGPT Pro tier, filling the gap between the $20 Plus and $200 tier. The plan focuses on significantly more Codex usage for heavy developer users, directly undercutting Anthropic and Google on pricing for power users.
Why it matters: This pricing move intensifies the subscription war among AI providers and could accelerate enterprise adoption of AI coding tools. The focus on Codex usage signals OpenAI sees developer tooling as a key monetization driver.
Claude Cowork Expands to All Paid Plans on macOS and Windows
Anthropic's Claude Cowork — the desktop agent that reads, edits, and creates files locally — is now available on all paid plans with new organizational controls and Zoom integration. Previously limited to Max subscribers on macOS only.
Why it matters: Computer-use AI agents are becoming mainstream enterprise tools. The expansion to Windows and all paid tiers dramatically increases Claude's addressable market for knowledge work automation.
Atlassian Launches Visual AI Tools and Third-Party Agents in Confluence
Confluence users can now create visual assets within the software, plus new third-party AI agents from Lovable, Replit, and Gamma work directly inside the platform.
Why it matters: Enterprise collaboration tools are rapidly becoming AI agent platforms. This integration pattern — where AI agents from multiple vendors work within existing tools — is likely the future of enterprise software.
Sierra's Bret Taylor: 'The Era of Clicking Buttons Is Over'
Sierra launched Ghostwriter, an agent that builds other agents. Users describe what they need in natural language, and Ghostwriter autonomously creates and deploys specialized agents to execute the task — replacing traditional click-based web applications.
Why it matters: The 'agent as a service' model could fundamentally change how enterprise software is built and consumed, shifting from pre-built UIs to on-demand, AI-generated workflows.
AI-Powered Intelligent Mobility Becomes Default Urban Operating Model
A new Miovision/451 Research report finds that AI-powered intelligent mobility — leveraging sensors, V2X communication, and generative AI — is positioned to become the default operating model for urban transportation. Advanced traffic management systems using AI can predict congestion, optimize signal timing, and proactively reroute traffic.
Why it matters: For parking and mobility operators, this signals that AI-driven traffic management and multimodal integration are moving from pilots to standard operations. Companies that don't adopt AI-powered optimization risk falling behind as cities mandate smarter infrastructure.
Z.AI Releases GLM-5.1: 754B Agentic Model with 8-Hour Autonomous Execution
Z.AI released GLM-5.1, a 754B MoE model that achieves state-of-the-art on SWE-Bench Pro (58.4), outperforming GPT-5.4 and Claude Opus 4.6. The model can work autonomously on a single task for up to 8 hours, running experiments, analyzing results, and iterating — a first for open-weight models.
Why it matters: Long-horizon autonomous AI agents are here. An open-weight model that can sustain 8 hours of autonomous engineering work changes the economics of software development and signals a shift from 'chat' to 'delegate' paradigms.
Google Gemini Adds Interactive Visualizations in Chat
Google Gemini now generates interactive data visualizations directly in the chat interface that users can tweak and explore, following a similar feature from Anthropic's Claude.
Why it matters: The chatbot interface is evolving beyond text into interactive, explorable content — raising the bar for what users expect from AI assistants.
Google and Intel Deepen AI Infrastructure Partnership
Google and Intel are co-developing custom chips amid a growing global GPU shortage. The partnership deepens their collaboration on CPUs, GPUs, and IPUs for AI infrastructure.
Why it matters: The chip shortage is reshaping AI industry alliances. Co-development deals signal that major cloud providers are diversifying beyond NVIDIA to secure compute capacity.
Florida AG Investigates OpenAI Over Shooting Linked to ChatGPT
Florida's Attorney General announced an investigation into OpenAI after ChatGPT was reportedly used to plan an attack at Florida State University that killed two and injured five. The family of one victim plans to sue OpenAI.
Why it matters: This investigation could set precedent for AI company liability in violent incidents. Regulatory scrutiny of AI safety is intensifying at the state level, with 19 new AI laws passed in 2026 alone.
19 New AI Bills Passed Into Law in 2026 — and Counting
Since mid-March, the number of new AI laws in 2026 has jumped from 6 to 25, with another 27 bills passing both chambers. Key legislation includes New York's RAISE Act for frontier AI developers, Washington's chatbot transparency requirements, and California's AI procurement executive order.
Why it matters: The patchwork of state AI regulations is accelerating rapidly. Enterprise AI deployments — including in parking and mobility — will need to account for an increasingly complex compliance landscape.
- → Anthropic's restricted Mythos release and the 19+ new AI laws highlight an accelerating regulatory environment. Enterprise teams should proactively assess AI compliance requirements across states where they operate, particularly New York's RAISE Act and Washington's transparency mandates.
- → Meta's Muse Spark with its multi-agent 'Contemplating' mode and Sierra's Ghostwriter agent-builder demonstrate that AI agents are moving from experimental to production. Automation teams should evaluate multi-agent architectures for complex operational workflows like dynamic pricing, resource optimization, and customer service.
- → The Miovision/451 Research report showing AI-powered intelligent mobility becoming the default urban model signals opportunity for operations-heavy industries to integrate real-time data systems with existing infrastructure.
- → Claude Cowork expanding to all paid plans on Windows means enterprise employees can now leverage AI desktop agents for document processing, reporting, and operational tasks at accessible price points — worth piloting across operations teams.
- → The AI pricing war (OpenAI's $100 Pro tier undercutting competitors) means enterprise AI tooling costs are dropping. Now is an optimal time to expand AI tool access beyond pilot groups.
Generated daily at 7:00 AM ET by Perplexity Computer · Feed: 63 articles from 7 sources