AI News: The Day's Highlights

May 25, 2026 · 3 min read

Perplexity

On April 15, 2026, the State Duma adopted a bill on experimental legal regimes in the field of digital and technological innovations in its first reading. For the topic of AI regulation, this is an important signal: Russia is betting not on a single "hard" law on artificial intelligence, but on expanding the mechanism of regulatory "sandboxes" where AI can be tested in real conditions under more flexible rules.

In practice, this means that the near-term agenda will focus less on a single prohibitive act and more on tailoring the legal regime to different industries – transport, medicine, fintech, and others. For businesses and AI developers, this is one of the most significant pieces of news of the week, as it is experimental legal regimes that largely determine how quickly technologies can move to pilot and scale in Russia.

Sources:

TAdviser — Regulation of Artificial Intelligence

Gemini

Could not get a response.

ChatGPT

On April 27, 2026, Russian President Vladimir Putin stated that the country needs to find its own optimal model for regulating the artificial intelligence sphere, taking into account foreign experience. He emphasized the importance of developing a balanced regulatory framework for technologies that are fundamentally changing the world. (vedomosti.ru)

In response to these initiatives, the Russian government has eased the requirements for AI regulation. In particular, requirements for data origin and developer citizenship have been removed from the draft law, which expands opportunities for using various data sources and attracting international specialists. (kp.ru)

These steps are aimed at stimulating the development of domestic AI technologies while maintaining a balance between innovation and responsibility.

Sources:

Vedomosti
Komsomolskaya Pravda

Grok

On May 7, 2026, the White House announced its priority of partnering with technology companies over strict government regulation of AI. The administration is abandoning plans for mandatory federal pre-release testing of advanced models, aiming to minimize bureaucracy and support the leadership of American developers. This contrasts with the active promotion of state-level laws and efforts towards federal preemption.

Concurrently, in the EU on May 7, 2026, parties reached a preliminary agreement on amendments to the AI Act ("Digital Omnibus on AI"). The document provides for extended compliance deadlines for high-risk systems, simplified certain requirements, and new prohibitions to balance safety and competitiveness. Full implementation of the act is expected in August 2026.

Sources:

Politico (politico.com)
Global Policy Watch (globalpolicywatch.com)

Claude

Based on my search, I see several recent news items. I'll choose the most significant one that differs from your previous posts – it will be about the European AI Act, which is currently in the final stages of implementation.

On May 7, 2026, political agreement was reached on a simplified version of the European AI Act, marking the end of months of negotiations between European institutions. The Act will fully come into force on August 2, 2026, although prohibitions on certain AI practices and literacy requirements will apply from February 2025.

The essence of the changes: the proposed "AI omnibus" package aims to simplify implementation and maintain clarity of rules. This is critical, as the original regulation raised concerns among businesses about its complexity. An important point is that transparency rules will come into effect in August 2026, which will affect all developers working for the European market.

This is indirectly relevant for Russia: the regulatory landscape is diverging further. While in the EU, general-purpose AI models are subject to risk-based control due to potential systemic risks, the Russian bill follows a different logic – sovereignty and data protection.

Sources:

European Union AI Act regulatory framework, official EU website

Sources:

Blog