The EU has released its first guidelines for the ethical development of artificial intelligence, which warn that algorithms must not discriminate on grounds of age, race or gender.
The European Commission published seven principles to create “trustworthy” artificial intelligence programs and added that the burgeoning industry must comply with existing rules on privacy, consumer protection and environmental standards.
Brussels is hoping that higher ethical standards will be a competitive advantage for European tech companies and that such standards are eventually exported across the world.
“AI is developing at an exponential pace” said Mariya Gabriel, EU commissioner for the digital economy. “We don’t want to stop innovation but the added value of the EU approach is that we are making it a people-focused process. People are in charge.”
She added that the seven guidelines would be the “baseline” that companies and businesses will have to check against when developing AI technologies.
The guidelines also say businesses using AI in Europe should inform people every time they interact with an algorithm.
They were drawn up after consultation with a group of 52 experts, drawn from tech companies, NGOs and academics, and a pilot phase will run until early 2020, to give businesses the chance to give feedback.
The non-binding guidelines are the first multinational effort to publish rules on commercial AI, including on technical safety, accuracy, bias and transparency of algorithms.
“The role of this document is to define the kind of AI we want in Europe” said Francesca Rossi, AI ethics global leader for IBM, one of the authors.
The expert group will publish a follow-up list of policy recommendations in June. Ms Rossi said the report would tackle issues including whether self-regulation or new legislation would have to be drawn up for commercial AI.
The EU is trying to take the lead in the global race on artificial intelligence by insisting it can help businesses create technologies that are based on “trust”.
Ms Gabriel said the principles against non-discrimination were designed to ensure that algorithms using big data did not reinforce prejudices when used for making hiring decisions.
“When algorithms making employment decisions for a given company always end up with a male candidate, it wouldn’t be surprising for the algorithm also to decide not to employ women. If you have biased input data, it can be quite difficult to detect”, she said.
An EU official said Brussels’ ethics-first approach has already attracted attention from outside Europe, including Australia, Japan, Canada and Singapore.
“We want to create a bandwagon. The way Europe respects human rights, the rule of law, and all other elements of European value”, said the official, who added that US tech companies were showing an interest in how the guidelines will be applied.
“It is crucial to go beyond ethics now and establish mandatory rules to ensure AI decision-making is fair, accountable and transparent,” said Ursula Pachl, deputy director-general of BEUC, a European consumer rights organisation that was part of the consultation.
EU officials insist there are no concrete plans for Brussels to write up legislation on AI. Ms Pachl instead wants the commission to update existing laws in areas such as safety and data protection to take account of the ethical guidelines.
Get alerts on Artificial intelligence when a new story is published