Trump Signs Executive Order on AI Safety — But Keeps It Voluntary
President Trump signed an executive order on Tuesday aimed at mitigating security threats posed by artificial intelligence, marking a notable shift from the administration's previously hands-off approach to the technology. The order was directly prompted by Anthropic's Mythos Preview model, which sent shockwaves through Washington in April when it demonstrated a superhuman ability to find critical vulnerabilities in the world's most-used operating systems and web browsers.
The headline provision: AI companies are asked to voluntarily submit their most powerful models for government testing up to 30 days before public release. Federal agencies have 60 days to develop the voluntary framework for securing frontier AI models. The order also establishes an AI cybersecurity clearinghouse to review and share information on vulnerabilities, working in coordination with the AI industry and critical infrastructure operators.
Crucially, the order explicitly bars the government from creating mandatory licensing or preclearance requirements for new AI models — making this a request, not a rule. The voluntary nature has drawn criticism from those who argue stronger guardrails are needed, especially given Mythos Preview's documented ability to identify over 23,000 potential vulnerabilities across more than 1,000 open-source software projects, with over 1,000 confirmed as high or critical severity.
Sources: NPR, CNBC, White House
Anthropic Files for IPO at $965 Billion Valuation
In what could become the largest AI IPO in history, Anthropic confidentially filed its S-1 prospectus with the SEC on Sunday, June 1. The filing comes less than a week after the company raised $65 billion in a Series H funding round that pushed its valuation to $965 billion — surpassing OpenAI's $852 billion valuation from March.
The numbers behind Anthropic's filing are staggering: the company announced in May that its revenue run rate has ballooned to $47 billion, up from $10 billion in annual revenue last year. A debut above the $1 trillion mark now looks like the base case if markets cooperate.
The IPO filing caps a remarkable period for Anthropic, which also hired Andrej Karpathy — the renowned AI researcher who co-founded OpenAI and led Tesla's Autopilot program — to head a new pre-training research team. Karpathy started at Anthropic in late May and is working on using Claude to accelerate pre-training research under team lead Nick Joseph.
As a confidential filing, detailed financial information won't be publicly available until Anthropic converts to a standard S-1. But Wall Street is already positioning for what many expect to be a landmark AI listing.
Sources: CNBC, TechCrunch, Anthropic
Microsoft Launches Its Own AI Models at Build 2026, Reducing OpenAI Dependence
At its annual Build developer conference on Tuesday, Microsoft unveiled MAI-Code-1-Flash — a 5-billion-parameter coding model built entirely in-house using clean, appropriately licensed data. It's a clear signal that Microsoft is building its own AI capabilities rather than relying solely on its partnership with OpenAI.
MAI-Code-1-Flash features adaptive thinking that stays concise for simple requests but allocates more reasoning budget for complex tasks. According to Microsoft's benchmarks, the model outperforms Claude Haiku 4.5 on price-to-performance across coding benchmarks and can solve harder coding tasks with up to 60% fewer tokens, achieving 85.8% on Microsoft's adversarial coding benchmark.
Microsoft also announced MAI-Thinking-1, its first in-house reasoning model, trained without any OpenAI data. Both models are rolling out to GitHub Copilot users across all paid tiers — Free, Pro ($10/month), Pro+ ($39/month), and Max.
The dual launch signals Microsoft's strategic shift toward model independence. While the OpenAI partnership remains valuable, Microsoft is clearly investing in having its own frontier capabilities, particularly for developer-focused workflows where it can integrate deeply with GitHub and VS Code.
Sources: Microsoft AI, CNBC, Microsoft Build Live Blog
Google Reimagines Search With AI Agents and Gemini 3.5 Flash
Google's sweeping Search overhaul — first announced at I/O 2026 in May — is now rolling out globally. Gemini 3.5 Flash is the new default model powering AI Mode in Google Search, generating output tokens four times faster than competing frontier models while beating Gemini 3.1 Pro on coding and multi-step agentic tasks.
The most significant change is the biggest redesign of the Google Search box in over 25 years. The new interface is fully AI-powered, dynamically expanding to understand user needs with intelligent suggestions that go far beyond traditional autocomplete. Users can now search across modalities — text, images, files, videos, and even Chrome tabs.
Perhaps most intriguing is Google's introduction of Search agents — information agents that operate in the background 24/7, intelligently reasoning across information for ongoing tasks. For complex projects like planning a wedding or managing a move, Search can build custom dashboards and trackers that users can continuously revisit. These agents will roll out this summer, starting with Google AI Pro and Ultra subscribers.
With Gemini 3.5 Pro expected later this month — CEO Sundar Pichai told the I/O audience "give us until next month to get it to you" — Google is positioning itself aggressively in both the consumer and developer AI markets.
Sources: Google Blog, Gemini 3.5 Announcement, TechCrunch