Latest news with #AIBoard
Euractiv
5 days ago
- Business
- Euractiv
The EU will not budge on deadline for generative AI rules
The Commission is under growing pressure to delay implementation of its landmark artificial intelligence law, the AI Act – but for generative AI models, also known as general-purpose AIs (GPAI), the legal deadline of 2 August is not expected to budge. The last weeks have seen several calls from EU countries and industry for the EU to "stop the clock" on the implementation of the AI Act – with the key complaint being that important guidelines, standards and compliance tools are not yet ready. Some of these appeals specifically target rules for GPAI models – meaning generative AI systems such as OpenAI's ChatGPT, or Mistral's LeChat – which are due to come into force soon on 2 August 2025. A group of some 1,000 lobbyists, academics, experts, and civil society representatives took part in drafting a Code of Practice for GPAIs, which is supposed to support compliance with the law. But the drafting process has taken longer than anticipated, with the group missing the original 2 May deadline. With less than a month to go before the GPAI rules kick in under the AI Act, the Code still needs to be adopted by the Commission – although Tech Commissioner Henna Virkkunen has confirmed it will be published before the 2 August deadline. In parallel, the AI Board a body comprising national representatives – has been discussing the possibility of a "grace period" for signatories to begin applying the Code, in light of the delays in finalising the document. Such a grace period would mean GPAI developers would not be expected to follow the code's specific provisions immediately. However, they would still be legally required to comply with the AI Act as of 2 August 2025. The AI Board is considering extending the deadline for applying the Code until the end of 2025, a Commission spokesperson told Euractiv . Any grace period would apply only to the voluntary Code of Practice and would not suspend the AI Act's GPAI rules themselves. "There is no stop the clock, there is no grace period, there is no pause," a Commission spokesperson said in a Friday press briefing. But it's worth noting that there is a year's compliance grace period baked into the Commission's draft guidelines on GPAI rules – which is why the AI Office, which oversees the AI Act, will not actually start enforcing the GPAI rules until 2 August 2026. For GPAI models already on the market before 2 August 2025, the rules will also only apply from 2 August 2027. Some lobbyists may be aiming for a broader delay in the AI Act's implementation. A Commission spokesperson told Euractiv that the rules on GPAI models will apply from 2 August. However, in the context of the upcoming digital simplification omnibus expected by year-end, they added that 'all options remain open for consideration at this stage.' (nl, aw)
Euronews
09-05-2025
- Business
- Euronews
Three months before deadline, EU countries not ready for AI oversight
With just under three months to go before the 27 EU member states need to appoint a regulator tasked with overseeing business' compliance with the AI Act, it remains unclear in at least half of the member states which authority will be nominated, a round-up by Euronews shows. By 2 August member states must notify the Commission about which market surveillance authorities are appointed, in addition the countries need to adopt an implementing law that sets out penalties and that empowers their authorities. The latest meeting of the AI Board in late March, which helps coordinate cooperation between member states, showed that the majority of the counties sent representatives from ministries. Just a handful - Denmark, Greece, Italy, Portugal and Romania - had national regulators attending. The EU executive does not want to comment on which countries are ready yet, but an official working at the AI Office told Euronews that the process in those member states that recently went through elections, such as Germany, will likely be delayed. The official said that the states are overall are having 'intense discussions' in the AI Board – which helps coordinate cooperation between member states – as there are different ways to set up the oversight structure. Countries are free to decide how to do it, and whether to appoint just one or several regulators. 'I think 95% of them have certainly chosen the structure that they want to have, and started the process to appoint the authorities. We will see whether on 2 August things will be finalized or not. Sometimes it's difficult to tell because the process in the parliaments may be more or less on here,' the official said. The AI Act – which aims to regulate AI tools according to the risk they pose to society – entered into force in August 2024, and started to apply gradually. It will be fully in force in 2027. A delay in appointing the oversight bodies will mean uncertainty for businesses that have to start complying with the rules. Some member states have set up an entirely new regulator, such as Spain, where AESIA, an independent agency of the Spanish Department of Digital Transformation, is likely to assume the role. In Poland, a pending implementing act sets up a new body, the Committee on Development and Security of AI, as the market surveillance authority. Denmark on the other hand designated its pre-existing Agency for Digital Government. For Germany, it seems likely that the Federal Network Agency will take up the role. Others, including the Netherlands, will likely expand the tasks of the privacy watchdog to also check compliance with the AI Act, which has the General Data Protection Regulation (GDPR) as its legal basis. The privacy regulators themselves called upon member states in July to ensure that they take charge of high-risk systems such as Biometric identification, law enforcement as well as migration, asylum and border control management.
