Advertisement
BusinessNews

EU AI Act Deadline Nears, But Code of Practice Won’t Arrive Until Year-End

Tech giants and regulators brace for a compliance crunch as Brussels delays key AI guidance

July 3 EST: Europe’s AI rulebook is nearly in effect—but the instruction manual is nowhere to be found. With just over a year before new transparency rules on general-purpose AI (GPAI) become binding under the EU AI Act, companies are still waiting on the Code of Practice meant to explain how to comply.

The European Commission now says the Code won’t arrive until late 2025, nearly seven months behind schedule. In boardrooms across tech and finance, that’s setting off alarms—not because of what’s in the Code, but because of what’s not clear without it.

Companies Are Flying Without a Flight Plan

The AI Act’s GPAI requirements—set to become law on August 2, 2025—ask developers to disclose how their models work, what data trained them, and whether they carry risk for bias, copyright violations, or excess power usage. It’s a disclosure-heavy rulebook with legal teeth, but no one’s handed out the playbook yet.

Big players like Google, Meta, Mistral, and ASML are asking the Commission to hit pause. Not forever—just long enough for the Code to come out, get digested, and start shaping real compliance strategy. “You wouldn’t enforce new building codes before publishing the blueprint,” one European policy advisor told Reuters.

The Commission’s response? The law is the law. The Code of Practice, they say, is optional—meant to offer clarity, not cover. But if you’re a firm looking at enforcement risk and thinking, “optional” feels a lot like “necessary,” you’re not alone.

Delay in Guidance, Not in Enforcement

This isn’t just about red tape. The Code was originally slated for May 2, 2025—months before the law locks in. But with delays, the Commission now says it’ll be out “in the coming days,” with industry sign-ons starting in August, aiming for a full rollout by year’s end.

Meanwhile, enforcement begins for new GPAI models in August 2026. Existing models have until August 2027 to comply. That two-year window might sound generous—until you realize what’s on the table: detailed audits of AI systems built on proprietary, often confidential, data stacks. Compliance isn’t plug-and-play. It’s forensic.

For startups, this is a red zone. “We don’t have the headcount to write policy memos and debug legal interpretations,” a CTO of a mid-stage AI firm in Berlin told Claims Journal. “We’re waiting on the Code to even start our risk audit.”

Voluntary in Name, Binding in Practice

On paper, the Code of Practice is voluntary. But as one Brussels-based tech lawyer put it: “The moment your competitor signs on and you don’t, you’re no longer just outside the club—you’re outside legal cover.” In absence of the Code, each company will have to interpret the AI Act’s vague sections themselves. That’s a recipe for uneven enforcement—and potentially, courtroom fallout.

And while the Commission says the Code will offer “legal certainty,” it hasn’t offered clarity on what happens to companies that don’t adopt it. The risk isn’t just fines. It’s inconsistent rulings from different national regulators, investor scrutiny, and a regulatory overhang that could slow deals or block launches.

This ambiguity is pushing even cautious firms to lobby for a “clock stop”—a delay in enforcement until guidance is available. The Commission has rejected that outright.

Europe’s Regulatory Gamble

The AI Act is the world’s first real attempt to regulate powerful AI systems. Washington is still debating frameworks; Beijing is selectively policing; Europe is betting on rules. But with this Code delay, it risks enforcing laws that no one can yet follow with confidence.

That’s not lost on EU member states. According to Reuters, several governments have joined industry calls to delay enforcement—not to water down the law, but to ensure companies can actually implement it. Without the Code, even basic decisions like how to disclose training data or measure model bias remain unsettled.

The worry isn’t just compliance. It’s competitiveness. Europe doesn’t want to become a “regulate-first, innovate-later” zone. But that’s the risk if mid-tier firms can’t keep pace with deep-pocketed multinationals who can afford to guess wrong and pay lawyers later.

In a vacuum of certainty, companies are improvising. Legal teams are mapping model disclosures to existing GDPR frameworks. Others are commissioning third-party audits as a hedge. But none of it guarantees compliance—and every step adds cost, delay, or exposure.

Until the Code of Practice is live, the Act’s transparency mandates are like a tax code without the IRS manual. You can try to comply. But you’re gambling on the fine print.


New Jersey Times Is Your Source: The Latest In PoliticsEntertainmentBusinessBreaking News, And Other News. Please Follow Us On FacebookInstagram, And Twitter To Receive Instantaneous Updates. Also Do Checkout Our Telegram Channel @Njtdotcom For Latest Updates.

A Wall Street veteran turned investigative journalist, Marcus brings over two decades of financial insight into boardrooms, IPOs, corporate chess games, and economic undercurrents. Known for asking uncomfortable questions in comfortable suits.
+ posts

A Wall Street veteran turned investigative journalist, Marcus brings over two decades of financial insight into boardrooms, IPOs, corporate chess games, and economic undercurrents. Known for asking uncomfortable questions in comfortable suits.

Source
ReutersReutersClaims Journal

Related Articles

Back to top button