EU and US set out plan to create rules of the road for artificial intelligence

At the third transatlantic Trade and Technology Council, Brussels and Washington agreed to draw up shared AI terminology and technical standards. But the two are still a way apart on how to regulate the technology.

The EU and US have set out a joint roadmap to find common ways to define and evaluate artificial intelligence (AI), though critics say they are still not going far enough to make sure AI protects democracy and human rights.

At the third Trade and Technology Council, (TTC) held outside Washington DC on 5 December, EU and US leaders announced joint initiatives on AI, digital infrastructure and manufacturing standards, although Brussels remains unhappy at what it sees as unfair subsidies for US electric vehicles.

They issued a first AI roadmap, which will “enable trustworthy AI systems that enhance innovation, lower barriers to trade, bolster market competition, operationalise common values, and protect the universal human rights and dignity of our citizens.”

Despite the talk of human rights, the roadmap is largely focused on getting the EU and US on the same page when it comes to the technical side of AI, and setting global technological standards, rather than banning certain worrisome uses like facial recognition in public spaces.

A working group will be set up try to pin down common definitions of key AI terms like trustworthy, risk, harm, bias, robustness, safety and so on.

The EU and US also want to provide “leadership” in setting international technical standards for AI, globally shaping rules on what counts as a reliable or accurate AI system, for example. They will also cooperate on “pre-standardisation” R&D to “advance the technical and scientific foundation for international standards development”.

Although China is not explicitly mentioned, their cooperation is likely to be seen as an attempt not to cede a standards setting role to Beijing.

There will also be a “knowledge-sharing mechanism” between the EU and US to give advance warning of the potential risks inherent in cutting-edge AI. And they will build a shared repository of metrics and methodologies for assessing whether an AI system is trustworthy or risky.

The AI roadmap “confirms our intention to set global standards for emerging technologies jointly” said Bernd Lange MEP, chair of the European Parliament’s trade committee, reacting to the announcement.

“As there are very different views on AI globally, it is crucial for the EU and the US to spell out clearly that all AI should be fully in line with human rights and democratic principles,” he said in a statement.

The joint roadmap is a “welcome initiative despite the differences between the AI governance approaches of the EU and the US,” said Risto Uuk, a policy researcher at the Future of Life Institute, which tries to steer potentially risky technologies.

But, he said, the roadmap’s objectives to set AI standards “lack concreteness” and the next steps “need to be made much more concrete and meaningful.”

No regulatory alignment, yet

Competition commissioner Margrethe Vestager said in late November that she wanted the TTC to “pave the way for a transatlantic sort of space for trustworthy AI” and that the US and EU had “very aligned approaches” when it comes to regulating the technology. Ultimately, companies on both sides of the Atlantic should end up being able to comply with both regulatory regimes using a “single set of tools,” she said.

But the latest TTC announcement leaves a long way to go before companies have anything like a level playing field in the EU and US when deploying AI systems. The roadmap is peppered with reminders that this joint work “does not constrain or prejudge the regulatory activities of the two parties.”

The EU’s AI Act is currently working its way through various levels of scrutiny by EU institutions, and should impose a legally binding framework on AI, outlawing some uses altogether.[…]