BRUSSELS, July 3 (Reuters) - A code of practice designed
to help thousands of companies comply with the European Union's
landmark artificial intelligence rules may only apply at the end
of 2025, the European Commission said on Thursday.
Alphabet's Google, Meta Platforms ( META ),
European companies such as Mistral and ASML as well as
several EU governments have called for a delay in implementing
the Artificial Intelligence Act, partly due to the lack of a
code of practice.
Publication of the Code of Practice for large language
models (GPAI), such as OpenAI's ChatGPT and similar models
launched by Google and Mistral, had originally been planned for
May 2.
The Commission plans to present the code in the coming days
and expects companies to sign up next month and the guidance
likely to kick in at the end of the year, a Commission
spokesperson said.
"On the AI Act's GPAI rules, the European AI Board is
discussing the timing to implement the Code of Practice, with
the end of 2025 being considered," he said.
Signing up to the code is voluntary, but companies who
decline to do so, as some Big Tech firms have indicated, will
not benefit from the legal certainty provided to a signatory.
The Commission pushed back against calls for a delay in
rolling out the AI rules.
"Our commitment to the goals of the AI Act, such as
establishing harmonised risk-based rules across the EU and
ensuring the safety of AI systems in the European market,
remains unchanged," the spokesperson said.
Campaign group Corporate Europe Observatory decried Big
Tech's role.
"Delay. Pause. Deregulate. That is Big Tech's lobby
playbook to fatally weaken rules that should protect us from
biased and unfair AI systems," said Bram Vranken, Corporate
Europe Observatory researcher and campaigner.
The AI rules for GPAI models will become legally binding
on August 2 but only enforced a year later for new models placed
on the market starting from next month. Existing models will
have two years to August 2, 2027 to comply with the rules.