AI News Roundup: White House Reverses on AI Oversight, Alphabet Nears World’s Most Valuable Company, EU Simplifies AI Act

White House Reverses Course on AI Oversight After Anthropic’s Mythos Sparks Cyber Alarms

In a dramatic policy reversal, the Trump administration is now weighing mandatory pre-release vetting for advanced AI models — a move driven by national security concerns over Anthropic’s Mythos model and its ability to autonomously discover and exploit software vulnerabilities at unprecedented scale. The White House is drafting an executive order that would create a formal government review process for frontier AI systems before they reach the public.

Mythos, which Anthropic has restricted to a select group of partners through its Project Glasswing cybersecurity initiative, has identified tens of thousands of zero-day vulnerabilities across every major operating system and web browser. Launch partners include Apple, Amazon, JPMorgan Chase, and Palo Alto Networks, with over 40 additional critical-infrastructure organizations granted access to scan and patch their systems.

The shift is striking given the administration’s earlier deregulatory stance. White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent recently met with Anthropic CEO Dario Amodei for what both sides called a productive conversation about government-industry collaboration — a thaw after months of tension following the Pentagon’s labeling of Anthropic as a national security concern.

Alphabet’s 160% Rally Puts Google Within Striking Distance of World’s Most Valuable Company

Alphabet briefly surpassed Nvidia by market cap in after-hours trading this week, closing the week at $4.8 trillion behind only Nvidia’s $5.2 trillion. The stock is up roughly 160% over the past year, driven by an emerging Wall Street consensus that Google owns more of the AI stack than any competitor — chips, models, infrastructure, and distribution.

The catalyst: a report that Anthropic committed to spend $200 billion on Google Cloud over five years for 5 gigawatts of compute capacity. That deal, combined with a cloud backlog that nearly doubled to $462 billion, prompted JPMorgan to call Alphabet its top overall pick in tech. Gene Munster of Deepwater Asset Management summed up the bull case: Google is one of the two best-positioned AI companies because they control most of the stack.

EU Agrees to Simplify AI Act Rules, Pushes High-Risk Deadlines to 2027–2028

The EU Council and European Parliament reached a provisional agreement on May 7 to streamline the AI Act as part of the Omnibus VII simplification package. The deal pushes application dates for high-risk AI systems to December 2027 for stand-alone systems and August 2028 for those embedded in products — giving developers significantly more runway to comply.

The agreement also extends regulatory exemptions previously available only to SMEs to include small mid-caps, reduces the grace period for transparency requirements on AI-generated content from six months to three months (new deadline: December 2026), and adds a new outright ban on AI used to generate non-consensual sexual content or child sexual abuse material.

NVIDIA Tops $40 Billion in AI Equity Bets, Led by $30B OpenAI Stake

NVIDIA has committed more than $40 billion in equity investments in just the first four months of 2026, cementing its transformation from chipmaker to full-stack AI kingmaker. The headline deal: a $30 billion stake in OpenAI. The remaining $10 billion-plus is spread across seven multi-billion-dollar public company deals and roughly two dozen private startup rounds.

Recent disclosed investments include up to $3.2 billion in Corning, the optical-fiber and ceramics supplier for AI data center fabric, and $2.1 billion in IREN, a data center operator pivoting from Bitcoin mining to GPU compute. CFO Colette Kress said NVIDIA invests where it needs to ensure compute capacity is built around its hardware — critics counter that the strategy creates a circular economy where customers spend NVIDIA’s own capital on NVIDIA chips.

India’s Supreme Court Launches AI Chatbot “Su Sahay” and Unified Court Data System

Chief Justice of India Surya Kant on Sunday announced the launch of Su Sahay, an AI-powered chatbot integrated into the Supreme Court’s website, alongside the “One Case One Data” initiative that will unify case information across high courts, district courts, and taluka courts into a single system. The chatbot, developed by the National Informatics Centre, will help citizens access case status, court procedures, and filing guidance without visiting courthouses.

The Chief Justice reiterated that AI will assist with administrative and research functions but cannot replace human judges in decision-making — a distinction that comes as the court separately addresses a growing problem of AI-generated fake case citations being submitted by lawyers.

Google Kills Project Mariner, Folds Browser Agent Tech into Gemini

Google quietly shut down Project Mariner, its experimental web-browsing AI agent, on May 4 without a public announcement. The browser-based agent, which launched to much fanfare, was discontinued as the industry shifts toward more reliable command-line and API-based agentic tools. Google said Mariner’s technology will live on inside the Gemini API and the new Gemini Agent product.

Connecticut Passes One of America’s Most Comprehensive AI Laws

Connecticut’s legislature passed SB5 on May 1, with Governor Ned Lamont expected to sign the bill into law. The AI Responsibility and Transparency Act covers employment-related AI decision-making, chatbot safety protections for minors, synthetic content labeling, frontier developer obligations, and a new state AI regulatory sandbox. Key provisions take effect October 1, 2026, making it one of the fastest timelines for compliance among state-level AI laws.