One month to go. August 2, 2026 is the pivotal date of the EU AI Act, the moment when Europe's artificial intelligence regulation moves from preparation to full enforcement. European businesses are arriving at this milestone in a paradoxical position: subject to some of the world's most detailed AI rules, yet still navigating genuine uncertainty on key points.
It's worth laying out exactly what kicks in, what remains in limbo, and what the current state of compliance actually looks like, because misinformation is circulating faster than clear guidance on this topic.
What Happens on August 2
From August 2, the transparency obligations under Article 50 become fully enforceable. Any operator deploying AI systems that interact with people must disclose that fact. Synthetic content must be made recognizable, with machine-readable labeling requirements arriving by December 2. The enforcement regime for general-purpose AI models, those underpinning major chatbots and large language model products, also activates on this date.
The penalties are significant. Fines reach 35 million euros or 7% of global annual turnover for prohibited practices, 15 million euros or 3% for violations of core obligations, and 7.5 million euros for providing false information to supervisory authorities.
The Digital Omnibus Limbo
Here is the point that most coverage glosses over. The Digital Omnibus package, proposed in November 2025 and receiving political agreement on May 7, 2026, would defer obligations for high-risk AI systems to December 2, 2027 for categories including biometrics, critical infrastructure, education, and employment, and to August 2, 2028 for systems embedded in products.
As of today, the text has not been published in the Official Journal of the European Union. In practical terms, that means the original August 2 deadlines remain formally valid. Legal observers broadly advise companies to prepare as if the August date is confirmed, treating any eventual deferral as a bonus rather than advance permission to pause compliance work.
How EU Member States Compare on AI Readiness
Italy is the first EU country to have a national framework fully aligned with the AI Act. Law 132/2025, in force since October 10, introduced among other things the criminal offence of spreading deepfakes. On June 10, the Council of Ministers approved implementing decrees in preliminary form, with final versions expected by October. Those decrees cover a ban on exclusively automated decisions in hiring and dismissal, mandatory AI training for healthcare and regulated professions, and limits on use in law enforcement and the judiciary.
The governance architecture is defined. Italy's National Cybersecurity Agency supervises the market, AgID handles notifications, and financial sector oversight sits with Banca d'Italia, Consob, and Ivass. For anyone operating in crypto, the parallel is direct: these are the same authorities responsible for MiCA supervision, and the underlying logic mirrors MiCA's structure: rules first, market access second, with compliance becoming a competitive differentiator.
Across the broader EU, the picture is uneven. Several larger member states, including Germany and France, are still working through national implementing legislation, meaning Italy's early alignment stands out. The European Commission's AI Office, established under the AI Act to coordinate enforcement, is expected to publish further guidance for providers of general-purpose AI models ahead of August.
One Billion Euros That Includes Web3
There is a funding dimension to this story that often gets overlooked. Article 23 of Italy's Law 132/2025 activates up to one billion euros through CDP Venture Capital for startups and SMEs, with over 300 million euros already allocated across more than 150 companies, according to CDP Venture Capital figures. Eligible sectors, listed explicitly in the legislation, include AI, cybersecurity, quantum computing, edge computing, and web3. That is a public capital channel that Italian startups working at the intersection of AI and blockchain, including those building AI agents and AI tokens, should be actively exploring.
Adoption Is Outpacing Skills
Italian companies with at least 10 employees using at least one AI technology. Source: ISTAT, December 2025
The Real Problem Isn't the Regulation
ISTAT data from December 2025 shows AI adoption among Italian firms with ten or more employees doubled in a single year, rising from 8.2% to 16.4%. But the same data reveals a structural gap. Nearly 60% of companies that considered AI investment decided against it due to a lack of internal skills, according to ISTAT. And more than one in four SMEs currently using AI cannot identify a specific business function it serves.
That is the real exposure. A regulatory framework that is technically advanced sitting on top of an adoption base that is operationally immature. Compliance risks becoming a paperwork exercise for consultants rather than genuine governance. Companies using AI in any form should arrive at August 2 having done three concrete things: map which AI systems are already in use, including those that entered quietly through third-party vendor software, classify each by risk level, and document who makes decisions about their deployment. Official guidance remains available on the European Commission's AI policy portal and on the Italian Department for Digital Transformation website.
The August 2 date is now four weeks away. For companies still treating AI Act compliance as a future concern, the calendar has run out. The practical checklist is short, but the window to complete it without scrambling is shorter.
