With less than a month to go before parts of the European Union's AI Act come into force, companies are calling for a pause in the provisions and getting support from some politicians.

Groups representing big U.S. tech companies such as Google owner Alphabet and Facebook owner Meta, and European companies such as Mistral and ASML have urged the European Commission to delay the AI Act by years.

The rules for general purpose AI (GPAI) models take effect on Aug. 2, a Commission spokesperson reiterated, adding that the powers for enforcing those rules start only on August 2 2026.

Under the landmark act that was passed a year earlier after intense debate between EU countries, its provisions would come into effect in a staggered manner over several years.

Some important provisions, including rules for foundation models like those made by Google, Mistral and OpenAI, will be subject to transparency requirements such as drawing up technical documentation, complying with EU copyright law and providing detailed summaries about the content used for algorithm training.

The companies will also need to test for bias, toxicity, and robustness before launching.

AI models classed as posing a systemic risk and high-impact GPAI will have to conduct model evaluations, assess and mitigate risks, conduct adversarial testing, report to the European Commission on serious incidents and provide information on their energy efficiency.

For AI companies, the enforcement of the act means additional costs for compliance. And for ones that make AI models, the requirements are tougher.

But companies are also unsure how to comply with the rules as there are no guidelines yet. The AI Code of Practice, a guidance document to help AI developers to comply with the act, missed its publication date of May 2.

"To address the uncertainty this situation is creating, we urge the Commission to propose a two-year 'clock-stop' on the AI Act before key obligations enter into force," said an open letter published on Thursday by a group of 45 European companies.

It also called for simplification of the new rules.

A Commission spokesperson said the European AI Board is discussing the timing to implement the Code of Practice, with the end of 2025 being considered.

Another concern is that the act may stifle innovation, particularly in Europe where companies have smaller compliance teams than their U.S. counterparts.

While the Commission is set for GPAI rules to come in force from next month, its plan to publish key guidance to help thousands of companies to comply with the AI rules by year end would mark a six-month delay from its May deadline.

EU tech chief Henna Virkkunen had earlier promised to publish the AI Code of Practice before next month.

Some political leaders, such as Swedish Prime Minister Ulf Kristersson, have also called the AI rules "confusing" and asked the EU to pause the act.

"A bold 'stop-the-clock' intervention is urgently needed to give AI developers and deployers legal certainty, as long as necessary standards remain unavailable or delayed," tech lobbying group CCIA Europe said.