Debate grows over a “right to code” as AI rules expand in EU and US

In short: As AI regulation grows, more policymakers and researchers are debating whether people should have a protected ability to write, share, and run AI code.

What's going on

The phrase “right to code” is showing up more often in AI policy debates. It is not a single, settled legal right. In practice, it is shorthand for protecting lawful access to computing power, and the ability to build and modify software and AI models.

In the European Union, the new AI Act is the first broad AI law. It uses a risk-based approach, meaning the stricter the potential harm, the stricter the rules (like how medicine with bigger side effects gets tighter controls). The law bans some uses, sets transparency rules for generative AI (tools that create text, images, or audio), and applies different obligations depending on whether a system is “high-risk” or a general-purpose model.

In the United States, there is still no single federal AI law. Rules come from existing areas like privacy, civil rights, consumer protection, labor, communications, and copyright. States are also acting, including Montana’s “Right to Compute” law, which limits when the government can restrict private ownership or lawful use of computing resources, and also asks for risk management policies in critical infrastructure that uses AI.

Legal scholars also note a complication: with modern generative AI, behavior often comes from training on large amounts of data, not from human-written instructions you can easily read. That makes “just inspect the code” less useful than it is for traditional software.

What to watch

Expect more arguments about where to draw the line between open development and safety rules. Supporters say openness helps competition and public oversight. Critics say the most powerful AI may need licensing, disclosures, and faster intervention when something goes wrong.

Source: NYTimes