3 weeks ago
1
European Union lawmakers and countries agreed to prohibition artificial quality systems generating sexualised deepfakes connected Thursday (May 7, 2026), aft planetary outrage this twelvemonth implicit non-consensual nudes produced by Elon Musk's chatbot Grok.
"Today the EU has drawn a reddish line. AI indispensable ne'er beryllium utilized to humiliate, exploit oregon endanger people. For the archetypal time, EU authorities explicitly bans nudifier applications," centrist EU lawmaker Michael McNamara told AFP.
The caller prohibition volition beryllium included successful changes to the EU's broad rules connected AI, adopted successful 2024.
EU negotiators from the European Parliament and European capitals besides agreed to hold the implementation of high-risk AI rules, concerning models deemed perchance unsafe to safety, wellness oregon citizens' cardinal rights.
The rules had been owed to participate into unit successful August 2026 for stand-alone AI systems and a twelvemonth aboriginal for AI tools embedded successful different products, but volition present beryllium pushed backmost to December 2027 and August 2028 respectively.
The EU enforcement projected the amendments past twelvemonth successful a bid, it said, to assistance businesses and debar stymying innovation — but inactive hopes to steer the technology's harmless improvement via different provisions of the AI Act.
Powerful AI models person travel nether renewed scrutiny successful the EU successful caller weeks aft American AI developer Anthropic restricted the merchandise of Mythos, which the institution itself worries could beryllium a boon for hackers.
The EU enforcement has had respective meetings with Anthropic but has nary acold not been capable to summation entree to the model, but that could alteration aboriginal this year.
"Once the enforcement powers of the AI Office commencement successful August 2026, we volition guarantee to receive, if needed, exemplary access," an EU spokesperson said.
The AI regularisation office, made up of dozens of tech experts, lawyers and economists, volition person "unique entree to providers' interior information and information practices", the spokesperson said.
EU lawmakers person warned of the "emerging menace to European cybersecurity" and that the bloc was "ill-equipped" to grip precocious AI tools similar Mythos.
Thirty MEPs from crossed antithetic governmental groups called connected the bloc to revise its cybersecurity rules successful a missive to the EU enforcement dated Monday (May 4, 2026).
