In AI development circles, the share of work where AI builds AI is finally being discussed in concrete numbers. Anthropic disclosed that Claude itself leads roughly a quarter of the work of building its next-generation model. Meanwhile, documents in the AI copyright lawsuit surrounding Microsoft have surfaced an executive's remark referencing 'astonishing theft.' Pressure around AI is tightening on both the investment and regulatory fronts at the same time.

Anthropic: Claude Leads a Quarter of the Work on Its Next-Gen Model

According to Business Times, Anthropic said Claude leads a quarter of the work of building its next AI model — meaning the company's own product is actually shouldering the core of the development work on its successor.

What matters here is that a concrete figure — one quarter — was put on the record. 'AI building AI' recursive improvement has long been discussed mostly as a slogan or a forecast, but it is now being stated as an actual measured share of the work. Taken together with the shift of Claude Code toward parallel agents that we covered previously, it looks like the application of AI to development work is steadily deepening.

Crusoe Raises $3.9 Billion at a $30.9 Billion Valuation to Build 'AI Factories'

Crusoe, a major AI-focused data center company, has raised $3.9 billion, bringing its valuation in this round to $30.9 billion. The funds will go not only toward massive data centers but also toward building small, assemblable modular 'AI factories.'

With power and grid constraints becoming visible in country after country, the move reads as a strategy to expand compute on two fronts: concentrated investment in giant facilities and small distributed deployments. As long as generative AI demand keeps outstripping supply, the infrastructure investment race seems set to heat up even further.

US FAA to Put $875 Million of AI Behind Air Traffic Control

The US Federal Aviation Administration (FAA) plans to invest $875 million in AI-based software to support air traffic controllers' work. TechCrunch likened controllers to 'crossing guards of America's skies,' framing it as AI applied to a job with an extremely heavy decision-making burden.

This stands out as a case of AI adoption in mission-critical public infrastructure getting underway with a concrete budget attached. Precisely because safety is the top priority in this domain, how it performs in actual operation will likely determine whether it spreads to other public sectors.

Microsoft AI Copyright Lawsuit: Documented Reference to 'Astonishing Theft' Emerges

The Orlando Sentinel reports that filings in an AI copyright lawsuit reveal that a Microsoft executive made statements amounting to an acknowledgment of something described as 'astonishing theft.'

Whether a litigant's own words are preserved in the documents is a factor that can heavily sway the outcome of a copyright case. It is a development suggesting that the debate over training data is entering the stage of 'what did they know' rather than 'what did they do.'

King Charles Hosts an AI Summit as Bengio Senses a Regulatory 'Pivot Moment'

An AI summit hosted by UK King Charles was held, with industry leaders including NVIDIA CEO Jensen Huang and Google DeepMind head Demis Hassabis in attendance. The King warned of the risk of catastrophic harm if the technology is misused, and executives from the participating companies also stressed the need for safety assurance, openness, and coordinated action.

Around the same time, Yoshua Bengio, one of the 'godfathers of AI,' suggested that technology regulation is approaching a pivot moment in government response, comparable to the one seen during the COVID-19 pandemic. Stuart Russell likewise argued that 'AI safety requires more than just slowing down.'

The debate over safety and regulation is shifting from warnings issued by individual researchers to a stage where the Royal Family and governments are providing the venue and drawing in the entire industry.

Elsewhere: Small LLMs and Agent 'Manners'

TechCrunch spotlighted PrismML, a small-LLM developer, reporting that 'small LLMs could change how we use AI.' The company's 'Ternary Bonsai 2 27B' is a talking point in Reddit's LocalLLaMA community.

Resy, a restaurant reservation service, suspended the account of a venture capitalist who attempted to book tables with an AI agent. It is drawing attention as an early case where a clash between a service and AI agents was documented in plain sight.

Wrap-Up

AI building AI, mega-scale infrastructure investment, real-world deployment in the public sector, and political pressure over safety — stories pointing in different directions all landed on the same day. The document release in the Microsoft lawsuit and the King-hosted summit, in particular, may mirror a turning point at which the conversation around AI moves from 'building' to 'responsibility.' We'll keep following how this develops.