Summary:
Barbie, en partenariat avec OpenAI, s’apprête à incarner un système d’IA à haut risque. En tant que jouet régulé par la Directive sur la sécurité des jouets, ChatGPT Barbie doit respecter des normes strictes de sécurité et de protection des données, notamment pour les mineurs. Mattel devra gérer la gouvernance des données, la cybersécurité et évaluer les risques associés. Si Barbie répond à des questions délicates, un mécanisme de rétrogradation sera mis en place pour assurer des réponses appropriées.
Original Link:
Original Article:
🤖🪆Barbie’s conquered the Dreamhouse, the courtroom, and the moon.
Next up: the EU’s high-risk AI registry.
As reported by Luiza Jarovsky, top brands Mattel and OpenAI “announced a strategic collaboration to support AI-powered products and experiences based on Mattel’s brands.” Read: Barbie’s going to be ChatGPT-powered.
And yes, that means Barbie’s going to be a high-risk AI system.
🧠 AI used for safety in Annex I regulated products is high-risk (art. 6(1) AIA). But that also goes for AI that qualifies as such a product itself. And ChatGPT Barbie is a ‘toy’ under the Toy Safety Safety Directive (which is in Annex I) and the upcoming Toy Safety Regulation. The chatty component is an inherent part of the toy.
🧠 But is that enough? As we recently debated, to qualify under Annex I we also need a third-party conformity assessment. For toys, that’s the general rule. The recent debate was about whether it matters that manufacturers can do the CA in-house if they apply EU standards. A last-minute amendment to the TSR says it doesn’t – I’m eagerly waiting for the OJ publication to see if this made it.
🧠 Let’s say ChatGPT Barbie is a high-risk AI. Then who is the provider? Likely not OpenAI, but Mattel, if it fine-tunes ChatGPT for child-safe responses (“It’s not a Dreamhouse—it’s an interactive learning environment!”)
That puts Mattel on the hook for:
🔐 Data governance (especially for minors)
🛡️ Cybersecurity
📊 Risk management with documented mitigation under harmonised standards
Imagine the risk assessment?
Risk: Barbie is asked about late-stage capitalism
Mitigation: System fallback to “Let’s go shopping!”