AI News Roundup: Google Eyes Orbital Data Centers, OpenAI Launches $4B Consulting Arm, Trump Admin Reverses on AI Oversight

Google and SpaceX in Talks to Put AI Data Centers in Orbit

Google and SpaceX are in preliminary discussions to build data centers in orbit, according to a Wall Street Journal report published Monday. The talks center on Google’s Project Suncatcher, a research initiative exploring solar-powered satellite constellations equipped with Google’s TPU chips and free-space optical links. Google plans to launch two prototype satellites by early 2027 in partnership with Planet.

The timing is strategic for SpaceX, which is pitching orbital compute as a core selling point ahead of its anticipated $1.75 trillion IPO later this year. Elon Musk has argued that space-based data centers will eventually be cheaper to operate than terrestrial ones, though analysts note that satellite construction and launch costs still make the economics challenging today. Google is also reportedly speaking with other launch providers.

The announcement follows Anthropic’s own deal with SpaceX last week to use computing resources from xAI’s Memphis data center, with both companies exploring potential orbital collaboration as well.

OpenAI Launches $4 Billion Deployment Company, Acquires Tomoro

OpenAI unveiled the OpenAI Deployment Company on Sunday, a new venture designed to help enterprises build and deploy AI systems at scale. The company is backed by a consortium of 19 global investment firms, consultancies, and system integrators, led by TPG with Advent, Bain Capital, and Brookfield as co-lead founding partners.

Alongside the launch, OpenAI announced the acquisition of Tomoro, an Edinburgh-based applied AI consulting firm with deep experience building real-time AI systems for clients like Tesco, Virgin Atlantic, and Supercell. The deal brings approximately 150 Forward Deployed Engineers to the new entity from day one. OpenAI revenue chief Sarah Dresser called the move a response to enterprise AI adoption reaching “a tipping point,” noting that organizations increasingly need hands-on deployment help rather than API documentation.

Trump Administration Reverses Course on AI Oversight

The White House is now actively considering mandatory pre-release testing for advanced AI models — a significant reversal from the Trump administration’s initial rejection of Biden-era AI regulations. The catalyst: Anthropic’s Mythos model, which demonstrated the ability to identify and exploit critical cybersecurity vulnerabilities in legacy systems, including decades-old bugs in financial infrastructure.

National Economic Council Director Kevin Hassett outlined the emerging framework, describing a potential executive order that would create a government-industry working group for frontier AI evaluation — something he compared to an “FDA for AI.” The Commerce Department’s Center for AI Standards and Innovation (CAISI) has already completed over 40 pre-release model assessments in partnerships with Google, Microsoft, and xAI.

The policy pivot is driven entirely by national security concerns rather than the ethical considerations that animated Biden-era regulation. Anthropic itself occupies an awkward position: its Mythos model triggered the urgency, yet the company was simultaneously labeled a supply-chain risk by the Pentagon after restricting military access.

Google Unveils Gemini Intelligence for Android 17

At the Android Show I/O Edition on Monday, Google previewed Gemini Intelligence — a suite of agentic AI features that transforms Android from an operating system into what the company calls an “intelligence system.” The standout capabilities include autonomous task completion, on-screen context awareness, and cross-app workflow orchestration.

Key features include Rambler, a voice tool that polishes dictated messages into coherent text; Create My Widget, which lets users build custom home screen widgets by describing them in natural language; and Chrome Auto Browse, which can autonomously book appointments and reserve parking spots. Google demonstrated Gemini reading a shopping list from Notes and automatically building a delivery cart across apps. The features will roll out on select Samsung and Google phones this summer, with broader availability later this year.

OpenAI Grants EU Access to GPT-5.5-Cyber as Anthropic Holds Back

OpenAI committed on May 11 to provide the European Commission access to GPT-5.5-Cyber, a specialized variant of its latest model designed for authorized security work including vulnerability identification, malware analysis, and patch validation. Under OpenAI’s Trusted Access for Cyber (TAC) program, verified defenders gain access to versions of GPT-5.5 with fewer restrictions than the public model.

The move creates a visible contrast with Anthropic, which has limited Mythos access to roughly 40 organizations through its Project Glasswing initiative. The European Commission confirmed it has held multiple meetings with Anthropic, but discussions are “not yet at the same stage” as the arrangement with OpenAI. The divergence highlights a growing strategic split: OpenAI is betting on broad institutional access to build trust, while Anthropic maintains tighter controls over its most capable models.

Colorado Passes Compromise AI Law After Two-Year Fight

Colorado’s legislature passed SB 189 late Monday, ending a fierce two-year debate over how to regulate AI in high-stakes decisions like hiring, lending, and housing. The compromise bill significantly waters down the original 2024 law (SB 205): instead of requiring companies to disclose how their AI systems work, they now only need to notify consumers when AI is being used and provide appeal opportunities.

The bill passed with broad bipartisan support — 57–6 in the House and 34–1 in the Senate — but reactions were mixed. Senate Majority Leader Robert Rodriguez acknowledged the trade-offs, while critics argued the law “does nowhere near enough” to protect consumers from opaque algorithmic decisions. Implementation has been pushed to January 2027. Meanwhile, Connecticut’s comprehensive SB5 is awaiting Governor Lamont’s signature, and Iowa signed a new chatbot safety bill into law.