The Signal — June 25, 2026
The AI industry is simultaneously trying to build shared rules and fighting over whether those rules were broken to begin with. OpenAI backs a new standards body while facing its largest copyright lawsuit yet from local newspapers, and Google keeps losing top researchers to rivals.
OpenAI and Linux Foundation Launch Appia Foundation for AI Standards
OpenAI and the Linux Foundation have launched the Appia Foundation, a new initiative aimed at creating standardized conformity specifications across the AI value chain. The foundation will work on common evaluation frameworks and safety practices for AI development.
The move signals a shift toward industry-led governance at a time when regulatory frameworks remain fragmented across jurisdictions. Rather than waiting for governments to impose rules piecemeal, OpenAI framed its involvement as supporting "shared standards for advanced AI," language that positions the company as a willing participant in oversight rather than a target of it.
Whether the Appia Foundation produces meaningful standards or becomes another industry body that diffuses regulatory pressure remains to be seen. The Linux Foundation brings credibility from its track record with open-source software standards, but AI safety evaluation is a far more contested and less settled domain. The practical test will be whether the resulting specifications carry any enforcement weight or simply codify what leading labs were already doing.
Sources: OpenAI Blog · Linux Foundation
Nearly 400 Local Newspapers Sue OpenAI and Microsoft Over Copyright
Approximately 400 local and regional newspaper publishers filed suit against OpenAI and Microsoft in the Southern District of New York on June 24, alleging the companies copied millions of copyrighted articles to train ChatGPT and Copilot without permission or payment. It is the largest coalition of local newspaper publishers yet assembled in the AI copyright fight.
The lawsuit follows similar actions by the New York Times and other major outlets, but the sheer number of plaintiffs here tells a different story. These are papers that lack the resources to negotiate individual licensing deals (the kind of publications AI companies are unlikely to approach with checkbooks). Their content, covering local government and community news, forms a significant chunk of the training data that makes large language models useful for everyday queries.
The case will likely turn on whether training on copyrighted material constitutes fair use, the same question at the center of every pending AI copyright suit. But the scale of this coalition could influence how courts weigh the market-harm argument — 400 papers collectively represent a substantial economic interest that's harder to dismiss than a single plaintiff.
Sources: Bloomberg Law · Windows Forum · Digg
Top AI Researchers Jonas Adler and Alexander Pritzel Leave Google for Anthropic
AI researchers Jonas Adler and Alexander Pritzel left Google for Anthropic, continuing a pattern of high-profile talent departures that has seen some of the company's most accomplished scientists leave. Previous exits include Noam Shazeer, who returned to Google briefly before leaving again for OpenAI, and John Jumper of AlphaFold fame, who joined Anthropic.
The departures matter beyond the individual names. Google built much of the foundational research behind modern AI — the Transformer architecture came from Google Brain — but has struggled to retain the people who did that work. Anthropic, founded by former OpenAI researchers, has become a particularly frequent destination, putting pressure on Alphabet's stock as investors question whether the company can maintain its research edge.
The talent drain doesn't mean Google is falling behind in capabilities. It still has massive infrastructure and data advantages. But consistently losing senior researchers to competitors suggests something about compensation or the ability to ship research into products that Google hasn't yet fixed.
Sources: TechCrunch · CNBC
Editor's Desk
We cut five stories today. The EU AI Act delay coverage and the Netherlands/MATCH Act/ASML story were both more than three weeks old. NVIDIA's Nemotron coverage was recycled product material without a fresh angle. Superhuman's acquisition of GPTZero didn't carry enough significance for a three-story edition. And the A24/Google DeepMind backlash was more of a culture-commentary piece than a developments story. Three stories today. We'll have more when the news warrants it.
