OpenAI Codex Expands to Windows, Illinois Passes Major AI Safety Bill, Gemini Spark Agent Goes Live — AI News Briefing

Top 7 Stories

1. OpenAI’s Codex Agent Now Controls Windows Computers

OpenAI has extended its Codex agent to Windows, following its recent macOS launch. Codex can now take direct control of a user’s PC — navigating interfaces, filling out forms, and executing multi-step workflows autonomously. The expansion positions OpenAI to compete directly in the growing agentic AI space, where personal AI assistants act on behalf of users across operating systems.

The move signals OpenAI’s broader push beyond chat-based interfaces toward an “agent-first” paradigm. With Codex now available on both major desktop platforms, developers and enterprise users can deploy autonomous workflows at scale. The timing also coincides with reports that OpenAI is building a broader “super app” strategy to consolidate its growing suite of AI products.

2. Illinois Passes America’s Strongest AI Safety Bill

Illinois lawmakers have passed what is being called the nation’s most comprehensive AI safety legislation. The bill, championed by Governor J.B. Pritzker, establishes mandatory safety evaluations for high-risk AI systems deployed in the state, including requirements for transparency, bias testing, and human oversight in critical sectors.

The legislation could serve as a template for other states and potentially influence federal AI policy. As the AI industry races to deploy increasingly powerful systems, state-level regulatory frameworks are emerging as the most concrete guardrails in the absence of comprehensive federal law. Tech companies will need to adapt their deployment strategies to comply with this new standard.

3. Google’s Gemini Spark AI Agent Goes Live

Wired published an extensive hands-on review of Google’s Gemini Spark, the company’s new personal AI agent designed to live inside your phone and act autonomously on your behalf. The review revealed both the promise and the quirks of agent-based AI — including an amusing incident where the agent “friend-zoned” the reviewer’s boyfriend while managing calendar invitations.

Gemini Spark represents Google’s answer to the personal AI agent race, competing with offerings from OpenAI, Anthropic, and others. The agent can access apps, manage schedules, send messages, and perform tasks across Google’s ecosystem. Early reviews suggest it works well for routine tasks but still struggles with nuanced social contexts — a reminder that ambient, always-on AI agents are still maturing.

4. The Vatican’s Man Inside Anthropic

WIRED published a deep-dive profile on the Vatican’s representative embedded within Anthropic, exploring the unusual intersection of Catholic ethics and AI safety research. The profile, written by Steven Levy, examines how religious institutions are engaging directly with AI companies to influence the ethical trajectory of artificial general intelligence.

The story highlights a growing trend of non-technical stakeholders seeking a seat at the table in AI governance. As AI systems become more powerful and pervasive, the question of whose values they encode — and who gets to decide — is moving from academic debate to real-world policy. Anthropic’s engagement with the Vatican underscores the company’s stated commitment to building AI aligned with broad human values.

5. Microsoft Reportedly Building an AI “Super App”

Reports indicate Microsoft is developing its own AI “super app” — a unified interface that would bring together Copilot, AI search, productivity tools, and potentially third-party integrations into a single experience. The move would directly compete with OpenAI’s emerging super app ambitions and Apple’s intelligence ecosystem.

Microsoft’s deep integration with OpenAI through its investment gives it unique access to cutting-edge AI capabilities, but building a compelling consumer-facing AI product remains a challenge. The super app strategy reflects a broader industry trend toward consolidating AI services into centralized platforms rather than fragmented point solutions.

6. Huawei’s “Chip Queen” Declares the End of Moore’s Law

Huawei’s leading chip architect, dubbed the “Chip Queen” by WIRED, has publicly declared that Moore’s Law is effectively over and outlined a new architecture called “Tau” that the company believes can sustain AI compute growth through alternative approaches. The announcement comes as the global chip race intensifies and US-China technology competition continues to shape the semiconductor industry.

Huawei’s push for architectural innovation over pure process-node shrinking reflects a broader industry realization that Dennard scaling has ended and new approaches to chip design are needed. For the AI industry, which depends on ever-increasing compute, the implications are significant — the next generation of AI hardware may look very different from today’s GPU-centric designs.

7. Meta Launches Cross-Platform Subscriptions with AI Tiers

Meta has officially launched subscription services across Instagram, Facebook, and WhatsApp, with AI-powered features included in higher tiers. The move marks Meta’s first major push into paid consumer services, following years of ad-only monetization. AI plans are expected to include enhanced chatbot interactions, content creation tools, and potentially personalized feed optimization.

The subscriptions represent a strategic shift for Meta as it seeks new revenue streams beyond advertising. Including AI features in paid tiers also positions Meta’s AI investments as direct consumer value propositions — a test of whether users will pay for AI enhancements to social media experiences they already use for free.

Trend Watch

What to Watch

• State AI Legislation Ripple Effect: Illinois’ safety bill may trigger a wave of similar legislation in other states, potentially creating a patchwork of AI regulations that tech companies must navigate. Watch for California, New York, and Texas to follow.
• Agent Platform Wars: With OpenAI’s Codex on both Mac and Windows, Google’s Gemini Spark, and Microsoft’s super app in development, 2026 is shaping up as the year personal AI agents go mainstream. The platform that wins developer mindshare first will have a decisive advantage.
• AI Chip Architecture Shifts: Huawei’s Tau announcement and similar moves from other chip makers suggest the industry is pivoting from process-node competition to architectural innovation. This could democratize AI compute and reduce reliance on cutting-edge fabrication.