Association for Computing Machinery (ACM) Office of Public Policy -- ACM is the world's largest and oldest society of professionals in all aspects of computing.
1/7 The EU AI Act's Article 50 transparency rules sound solid in theory. But can they actually be enforced? @acmpolicy's Europe TPC says the draft guidelines have a technical problem. 🧵
7/7 Bottom line from @acmpolicy Europe TPC: Enforcement must be grounded in empirical detection limits & verifiable criteria — not assumptions about what "average users" notice. Full response: https://t.co/hshhh2EdnN
"Los responsables políticos deberían planificar la integración de estrategias avanzadas en el marco legal, en particular para hacer frente a los intensos picos de carga generados por el entrenamiento de grandes modelos de IA." https://t.co/4QrDGIpQDf #PolíticaTecnológica
La UE exige reportes de consumo energético a centros de datos, pero no fija límites vinculantes de eficiencia. Singapur: EUE ≤ 1,3. Australia: 1,4. California y China: 1,5. La UE: sin objetivo. #PolíticaTecnológica#IAResponsable
https://t.co/5dKlmAHHBi
7/7
"Advancements in government efficiency should not come at the expense of human dignity, autonomy, privacy, or safety."#TechPolicy#Privacy#DataPrivacy
Full statement: https://t.co/xDoyMjkI3m
1/7
🧵 NEW from @ACMPolicy: USTPC calls for an immediate halt to mass federal data consolidation efforts — merging income, health, benefits & immigration data across dozens of agencies into unified personal profiles. Here's why this matters. ⬇️
6/7
USTPC calls for: Independent technical & ethical review; Transparent disclosure of data sources & methodologies; Full Privacy Act compliance; Individual rights to access, correct & challenge data; Meaningful public oversight before deployment
EU data center sustainability framework proposing five recommendations: binding PUE targets, broader reporting coverage, stronger water-management obligations, renewable energy quotas, and long-term planning
https://t.co/eTp7mtqNIn
Lessons from Singapore's SEA-LION and Switzerland's Apertus: full pretraining from scratch is rarely the right starting point for resource-constrained governments. Open-source base models using national data works
https://t.co/0ilUe57uCB
As more states pass AI safety laws surrounding the use of chatbots, especially use by minors, we stand by our statement on the unethical development of these apps: https://t.co/K7mcqbuY5b