Governing AI: Ethics, Rules & Our Future
Debating ethical rules and global laws (like the EU AI Act) for advanced AI to ensure safety and align with human values.
Topics: technology, politics, individual and values
Example
The European Union's AI Act, agreed in December 2023, entered into force on 1 August 2024 as the first broad law regulating AI by level of risk. Its bans on 'unacceptable risk' uses, such as government social scoring and some manipulative systems, have applied since 2 February 2025. Strict rules for high-risk AI, for example in hiring, credit and policing, were due from August 2026, but in May 2026 EU lawmakers provisionally agreed to delay most of them to December 2027, saying technical standards were not ready.
Key facts
- The EU AI Act entered into force on 1 August 2024, and its bans on unacceptable-risk AI have applied since 2 February 2025.
Limitations
- The delay was a provisional agreement still needing formal adoption when this record was checked, and the Act's real effect depends on enforcement that had barely begun.
Evaluations
technology evaluation
Support
The Act sorts AI by risk, so a spam filter faces few rules while a hiring tool must be tested and documented, matching regulation to potential harm.
Counterargument
Firms complained the rules were unclear and costly, and the 2026 delay shows even the EU found the timetable too ambitious.
Rebuttal
Delay to get standards right is not abandonment; bans on the most harmful uses already apply.
Additional support
Rules for general-purpose AI models such as chatbots have applied since August 2025, showing the Act adapting to technology that arrived after drafting began.
politics evaluation
Support
Because many global firms will follow EU rules to sell in Europe, the Act may set standards far beyond the EU, as its data protection law did.
Counterargument
The US and China have not copied the EU approach, and the 2026 delay came under pressure to keep European firms competitive.
Rebuttal
Influence can still spread through trade even without copies of the law, but the delay shows competition concerns can slow regulation.
Additional support
The provisional deal also added a ban on AI that creates non-consensual sexual images, showing lawmakers responding to new harms.
individual and values evaluation
Support
Banning social scoring protects individuals from being judged by an opaque system that ranks their trustworthiness.
Counterargument
Some argue strict rules will slow medical and educational AI that could benefit individuals, so caution has its own costs.
Rebuttal
The risk-based approach tries to balance both values: freedom to innovate in low-risk areas and protection where rights are at stake.
Additional support
The Act gives people rights to explanations for some high-risk AI decisions, such as loan refusals, putting individual dignity into law.
Sources
- European Commission: AI Act: regulatory framework for AI
- Sidley Austin (Data Matters) (2026-06-22): EU Lawmakers Reach Provisional Agreement to Delay Key EU AI Act Obligations
More GP guides for this example
Topic guides
- Society and Culture GP Examples
- Politics and Governance GP Examples
- Science and Technology GP Examples