xAI supports AI safety and will be signing the EU AI Act’s Code of Practice Chapter on Safety and Security. While the AI Act and the Code have a portion that promotes AI safety, its other parts contain requirements that are profoundly detrimental to innovation and its copyright provisions are clearly over-reach.

Worth noting that the EU AI Act is still binding on Meta (as long as they sell in Europe). The Code is a voluntary compliance tool that provides a pre-approved pathway to demonstrate you're meeting the EU AI Act's requirements, but if you don't sign the Code you still need to demonstrate compliance in another way. Articles 53 and 55 of the EU AI Act still include specific requirements around detailed technical documentation for authorities, information sharing with downstream providers, copyright compliance policies, comprehensive safety and security frameworks, risk assessments, and incident reporting. (Likewise, xAI will still need to follow EU Copyright Law). These obligations kick in August 2, 2025 for new models and August 2027 for existing ones. Meta is essentially saying "we'll comply with the law, but we'll do it our own way." In my opinion this is actually a harder path - it's not really clear how one would demonstrate compliance through alternative means and Meta would face "more intensive scrutiny" from regulators. Meta's main objections are that the Code "introduces legal uncertainties" and goes "far beyond the scope of the AI Act" - basically arguing the voluntary guidance is more burdensome than the actual text of the EU AI Act itself. But Meta's refusal to sign the Code means they're betting they can demonstrate compliance more efficiently on their own terms and/or win in Court... a risky strategy that could backfire if regulators and the Courts decide their alternative approach falls short.