Illinois Passes the Strongest AI Safety Bill in the US
On May 27, the Illinois General Assembly passed a bill requiring major AI companies such as OpenAI, Anthropic, and Google to have independent third-party organizations verify their compliance with safety standards. According to WIRED, it is on track to become the most powerful AI safety law at the US state level. Governor JB Pritzker has indicated that he intends to sign it.
The bill's defining feature is its requirement for audits by independent third parties rather than corporate self-reporting. Enshrining in law that the safety of AI models must undergo external verification—not just assessment by the developers themselves—is a groundbreaking measure for US state law. Ahead of implementation, details such as the concrete criteria for audits and the scope of covered models remain to be worked out.
OpenAI Foundation Commits $250 Million
The OpenAI Foundation has announced that it will commit $250 million (roughly 39 billion yen) in initial funding to respond to the abrupt changes that the spread of AI will bring to labor markets and the economy. According to ITmedia, the funds will be allocated to three areas: "measuring economic impact," "supporting workers through transitions," and "building new models of economic stability."
Following an organizational restructuring, the foundation will be responsible for governance and the pursuit of public-interest objectives, and it will be closely watched for how it strikes the balance between the benefits and risks of AI development. A New York Times opinion piece has also fueled discussion of an "AI barons' giveback," adding momentum to the debate over how the AI industry's wealthy should contribute to society.
Cognition AI Raises $1 Billion at a $26 Billion Valuation
Cognition AI, the developer of the AI software engineer "Devin," has raised $1 billion at a valuation of approximately $26 billion (roughly 4 trillion yen). Since its unveiling nine months ago, Devin has attracted considerable attention as an agentic AI with autonomous coding capabilities.
The valuation is yet another sign of the fervent investment in the AI agent space. Beyond coding, the commercialization race for AI agents that can execute tasks autonomously is expected to grow even more intense.
AI Coding Agent Market Enters a "New Phase"
Gartner analyzes that, driven by advances in AI, the market for AI coding agents has entered a "new stage." According to ITmedia, the firm presents three reasons why integrated development environments (IDEs) are no longer indispensable.
This shift could change the very environment in which engineers work. Tools themselves are evolving in a direction that encourages independence from the IDE—for instance, Claude Code v2.1.152 adds a `/code-review --fix` feature that applies review results directly to code.
Google AI Can't Fix Spelling Mistakes in Image Generation
TechCrunch reports that Google's AI image generation cannot even spell the company's own name, "Google," correctly. Inaccurate text rendering is known as a structural challenge for diffusion models, but the fact that misspellings of its own brand go unaddressed is striking.
Text rendering accuracy in image-generating AI is an area where every provider is working on improvements, but a fundamental solution is expected to take more time.
---