DeepMind Exodus, Claude Tag Launches, OpenAI Splits With White House — AI News Briefing

Top 7 Stories

1. Google Stock Plunges as Top DeepMind Scientists and Nobel Laureate Defect to Rivals

Alphabet shares dropped as much as 6% after two of Google DeepMind’s most prominent researchers — including a co-author of the seminal Transformer paper and a Nobel laureate — defected to OpenAI and Anthropic. The departures mark an escalation in the AI talent war, with rivals poaching key architects of the technologies that underpin today’s generative AI boom. Investors wiped billions off Alphabet’s market cap as analysts questioned whether Google can retain the talent needed to lead in the AI race, especially as its Gemini model struggles to keep pace with frontier offerings from OpenAI and Anthropic.

2. Anthropic Launches Claude Tag — an Always-On AI Teammate Inside Slack

Anthropic unveiled Claude Tag, a persistent AI agent designed to live directly inside Slack workspaces. Unlike the previous Slack integration, Claude Tag learns from ongoing conversations, monitors channels autonomously, and can be summoned with @Claude Tag to execute tasks, answer questions, and track project status without needing a separate chat interface. The launch signals Anthropic’s push to embed AI agents directly into enterprise workflows as persistent “teammates” rather than on-demand assistants. Admins have until August 3 to migrate from the legacy Anthropic Slack app.

3. OpenAI Diverges From White House on AI Safety Rules

In a new policy paper, OpenAI broke ranks with the White House’s approach to AI safety regulation, arguing for a lighter-touch framework that emphasizes voluntary commitments over mandatory compliance. The divergence is significant: OpenAI had previously been a vocal supporter of government-led safety oversight. The shift reflects growing tensions between AI labs — including Anthropic and Meta — and the administration over how aggressively to regulate frontier models. Critics warn that voluntary frameworks could repeat the failures of social media self-regulation, while OpenAI argues that overly prescriptive rules could cede the AI race to China.

4. Nvidia Posts Record $58.3B Profit and Invests $40B in AI Startups This Year

Nvidia reported a record quarterly profit of $58.3 billion, driven by insatiable demand for its H200 and next-generation B200 AI accelerators. The company has also committed over $40 billion in equity investments to AI startups this year alone, cementing its role as both the dominant chip supplier and a major strategic investor in the AI ecosystem. Despite the earnings beat, Nvidia’s stock remains down ~18% year-to-date as investors rotate into smaller AI names and worry about long-term competition from AMD, custom chips by Broadcom, and in-house silicon at Google, Meta, and Microsoft.

5. Meta Under US Pressure to Submit AI Models for Government Security Reviews

The U.S. government is intensifying pressure on Meta to submit its largest AI models for mandatory security reviews, citing national security concerns. Meta has resisted, arguing that its models are open-weight and already subject to community auditing. The standoff echoes similar tensions between Washington and other AI labs, as the White House pushes for standardized red-teaming and pre-deployment testing. The New York Times reports that the administration is considering executive action if Meta does not voluntarily comply, raising the stakes for the broader debate over open-source AI governance.

6. Google and Meta Race to Build Personal AI Agents as OpenAI and Anthropic Pull Ahead

Both Google and Meta are accelerating development of personal AI agents — always-on, context-aware assistants that can act autonomously across apps and services — as they scramble to close the gap with OpenAI’s ChatGPT and Anthropic’s Claude. Google is reportedly integrating agentic capabilities directly into Gemini and Google Assistant, while Meta is building agent features into its messaging apps and smart glasses. Industry analysts note that the winner in the consumer AI agent race will likely control the next major computing paradigm, making this one of the highest-stakes product battles in tech.

7. Coinbase Launches AI Agent Accounts That Can Trade and Spend Cryptocurrency

Coinbase launched a new platform allowing AI agents to hold their own cryptocurrency wallets, execute trades, and make payments autonomously. The move marks a significant step toward AI-crypto convergence, where autonomous agents could manage digital assets, pay for API access, participate in DeFi protocols, and even negotiate with other agents — all without human intervention. Critics flagged security and alignment risks, warning that autonomous agent wallets could be exploited or act in unintended ways if not carefully constrained.

Trend Watch

What to Watch

• The Anthropic-Pentagon Standoff: Claude Tag’s enterprise push comes amid reports that Anthropic is in tense negotiations with the Pentagon over military uses of its AI, including red lines around autonomous weapons and mass surveillance. Another report suggests Anthropic may be walking back some of its core safety promises. Expect more news here this week.
• AI Safety Executive Order Watch: With OpenAI diverging from the White House and Meta refusing security reviews, the administration may turn to executive action. A new national AI policy framework (NSPM-11 follow-up) is reportedly in development.
• AI Agent Economy Goes Live: Between Coinbase’s agent wallets, Anthropic’s Claude Tag, and Microsoft’s pay-as-you-go AI agents, June 2026 may be remembered as the month the autonomous AI agent economy transitioned from concept to reality. Watch for first real-world agent-to-agent transactions.
• Nvidia GTC Taipei at COMPUTEX: Ongoing this week, Nvidia’s Taipei event could reveal next-generation architecture details and additional enterprise AI partnerships.
• House AI Preemption Draft: A new House draft bill would preempt state-level AI laws with federal standards — a move that could reshape how labs like Anthropic, OpenAI, and Meta navigate the regulatory patchwork.