All news

Europe's AI Act makes compliance part of the product

Graphite steel caliper closed around a white porcelain cube, a cobalt gauge block beside it

The European Parliament adopts the first comprehensive AI law. Model makers and high-risk uses now face dated duties, and proof of trust becomes a moat.

By 523 votes to 46, with 49 abstentions, the European Parliament adopted the Artificial Intelligence Act on 13 March, confirming the political deal struck with member states in December.

Europe's AI rules now come with dates, and dates become costs that investors can price into every company selling AI in the EU, a wider reach than the US order. Once the text is published, the bans apply after six months, the rules for general-purpose models after twelve, most of the law after twenty-four and the remaining high-risk obligations after thirty-six. In this market trust becomes part of the product, and diligence has to cover a company's risk class as carefully as its revenue.

The law reaches from the use case to the model. It bans a short list of practices, among them social scoring, emotion recognition in workplaces and schools, and untargeted scraping of facial images to build recognition databases, and breaches of the bans can cost up to €35 million or 7% of global annual turnover. It sets duties for high-risk uses in areas such as employment, education, banking, healthcare and critical infrastructure. Providers of general-purpose models must meet transparency rules, respect EU copyright law and publish detailed summaries of their training content, while the most powerful models face evaluations and incident reporting.

Fixed costs reward scale. Compliance favors those that can spread it, such as large model providers and established software vendors, and it creates demand for the documentation, testing and monitoring the law requires. That second effect is the opening. Fanaura, whose mandate covers the EU alongside the US and Latin America, looks for AI-native companies that turn the requirements into products, from evaluation and documentation to monitoring and data provenance, sold to the many firms that now need them.

A startup selling AI into hiring, education, banking or health in Europe should map its risk class now. The high-risk duties call for risk management, usage logs, human oversight and documentation, which are cheap to design in and expensive to retrofit. Teams that build on general-purpose models will need their providers' documentation and should ask for it in contracts, and every member state must set up regulatory sandboxes open to startups for testing before launch. Treated as a feature, with audit trails and clear data provenance, compliance shortens sales cycles with European enterprises rather than lengthening them.

Delayed launches would expose the cost. If companies hold back European releases of new models and features, the burden of compliance falls on European users. Two steps set the pace: formal endorsement by the Council and publication in the Official Journal start the clock, and the codes of practice for general-purpose models, due nine months after entry into force, will show how heavy the duties on model providers become in practice.

Europe has chosen to compete on trust rather than speed, and a law with a calendar turns trust into something a company can build, document and sell.

For founders

Building evaluation, documentation or monitoring tools for AI under the EU AI Act? Send us the project.

Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.

Apply in five minutes
Subscribe

Get the news by email

One short note each weekday. No marketing, and you can unsubscribe at any time.