News and Insights

Denmark’s culture minister introduced a bill on 8 October to ban sharing realistic digital imitations of a person’s appearance or voice without consent. Performers’ work is covered too, and satire stays allowed. The minister hopes it takes effect at the turn of the year. For anyone generating synthetic voices or faces, the question moves from whether output is labelled to whose likeness it uses, and who agreed to that.

AMD will substantially increase chip supply in 2027 and now plans three to five years ahead, chief executive Lisa Su told reporters in Taipei on 6 October. Demand still exceeds supply, AMD needs more advanced wafer capacity and Su called Taiwan critical to the chip supply chain. An AI roadmap set a year at a time quietly assumes that capacity will be there. How far ahead does yours look?

On 29 September the European Commission opened a consultation on how technology, AI included, affects copyright protection. The use of protected content in AI is one of four focus areas. Rights holders, generative AI providers and other AI value-chain actors are among those invited to respond until 3 November 2026. An organisation deploying generative AI arguably sits in that chain. Is its view on licensing better heard now, or after any new measures take shape?

BNP Paribas has signed a five-year Google Cloud partnership covering infrastructure, Gemini models and internal AI agents. It is not a wholesale move. The bank’s existing governance framework decides which data may run in public cloud, and certain categories will stay out. Each agent is to be limited to the resources its task needs. For an EU-operating organisation, the question is whether that boundary is drawn before the contract or after it.

The EDPB has adopted guidelines setting a five-step test for when a data protection authority should fine rather than reprimand. Step four carries the weight. A minor infringement generally draws a reprimand, while a non-minor one brings a strong presumption that a fine follows. For an EU-operating organisation, the live question moves from whether a fine is likely to whether its own breach would count as minor. Consultation runs to 13 November.

The Commission’s KIDS Act proposal, adopted on 17 September, would stop AI companions and chatbots reaching the market until providers can demonstrate compliance and run a monitoring mechanism for emerging risks. Enforcement rides on the AI Act for chatbots and the DSA for platforms, with fines up to 6% of worldwide turnover. For anyone shipping a chatbot that minors can reach, that is a market gate rather than a conduct rule.