Welcome to the EuropeanAI Newsletter covering the European AI and technology ecosystem. If you want to catch up on the archives, they're available here.

You can share the subscription link here. 

The French Presidency of the Council of the EU (starting 01/01/22) announced climate, social and digital as its mandate’s priorities. For digital it will investigate hate speech and the legislation on digital services and markets (DSA and DMA). Interestingly, there is no mention of the AI Act. 

The Council of the EU published its first amendments to the AI Act. Among those is an article on general purpose AI systems. However, the change itself doesn’t seem change much as general purpose AI system’s themselves would not fall under the regulation unless their intended purpose (when being put on the market) falls within the scope of the regulation and therefore makes them subject to the regulation. Presumably, this would have been applicable even before the amendment of the Council.

Dragoș Tudorache (Renew, Romania) and Brando Benifei (S&D, Italy) will be the EU Parliament’s negotiators for the AI Act.

Yesterday, the UK government’s Joint Committee on the Draft Online Safety Bill published its 190 pages long report on the issue. You can find a summary of recommendations entitled ‘Big Tech has failed its chance to self regulate’ here.

The European Commission has proposed new measures to improve the working conditions of workers in digital labour platforms. Among the proposals, Art 6 mandates the disclosure of the presence and characteristics of automated monitoring and decision-making systems used by the platform that could significantly impact working conditions to platform workers; Art 7 requires human monitoring of individual decisions involving automated systems to be undertaken by platforms, and for risks of automated systems on working conditions to be assessed regularly. Art 8 provides for the right of explanation and appeal to significant decisions supported by automated systems. It is unclear what “significant impact” on working conditions may mean when applied in practice. These provisions support the broader goal of the Directive to expand existing protections for personal data processing by automated decision-making systems present in the General Data Protection Regulation, and proposed requirements for AI systems established in the AI Act.

The Centre for Data Ethics and Innovation has released an independent report presenting a ‘Roadmap to an Effective AI Assurance Ecosystem’, describing the route towards an ecosystem of tools and services that support an ‘effective, pro-innovation governance of AI’. The roadmap identifies two challenges; the first being the ‘information problem’, describing the availability of evidence to determine whether an AI system is trustworthy, and the second is a ‘communication problem’, exploring how to inform users on a system’s trustworthiness. Their ambition is to have the assurance ecosystem up and running within the next 5 years.

Notes in the Margins: Inspiration for the UK’s assurance services can be found in the AI Act’s process to ensure that a high-risk AI system adheres to various legal obligations and in previous work by researchers, for example in this piece.