Direct comparison
Bletchley vs Seoul AI Safety Declarations
How the Bletchley (2023) and Seoul (2024) AI safety declarations differ, and how Seoul's company pledges became mandatory under SB 53 and the RAISE Act.
Written and maintained by CASRAI Editorial Board
Last updated
Ask CASRAI · free to try
Ask about Bletchley vs Seoul AI Safety Declarations
Ask your first 2 questions free below. Subscribers get 150 a day for $29 a month.
Ask CASRAI answers research-administration questions and cites the passages behind every claim. When our sources don't cover a question, it says so.
Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.
Works on this site and inside Claude, Cursor and the AI tools you already use.
Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays free to read, with no account and no card.
How do Bletchley Declaration (2023), Seoul Declaration / Frontier AI Safety Commitments (2024) compare side by side?
The table below compares Bletchley Declaration (2023), Seoul Declaration / Frontier AI Safety Commitments (2024) across 7 procurement-relevant dimensions, from date and location through how it feeds into binding regulation.
Side-by-side comparison
| Dimension | Bletchley Declaration (2023) | Seoul Declaration / Frontier AI Safety Commitments (2024) |
|---|---|---|
| Date and location | 1-2 November 2023, at Bletchley Park, UK, hosted by the UK government. | 21-22 May 2024, held virtually and co-hosted by South Korea and the UK, with the leaders' declaration and the industry commitments produced as two separate documents. |
| Who signed | 28 countries plus the EU, including the US, China, and the EU as a bloc — the first time China joined the US in a multilateral AI-safety statement. New Zealand added its signature in October 2024. | The Seoul Declaration (leaders' text) was signed by a narrower group than Bletchley's: Australia, Canada, the EU, France, Germany, Italy, Japan, South Korea, Singapore, the UK, and the US — 11 countries and entities. |
| What governments actually committed to | General, aspirational language: support an internationally inclusive research network on frontier AI safety, build risk-based domestic policies, and expect AI developers to provide “context-appropriate transparency and accountability” on measuring and mitigating harmful capabilities. No thresholds, mechanisms, or deadlines are specified. | Commitment to develop interoperable AI governance frameworks across countries, partly by integrating the Hiroshima Process International Code of Conduct, and to pursue human-centric AI development with industry, academia, and civil society. Similarly general — the leaders' text itself sets no thresholds or deadlines. |
| Company-level commitments | None — Bletchley produced no separate industry document. | A distinct “Frontier AI Safety Commitments” text, signed by 16 AI companies at the time (including Amazon, Anthropic, Google, Microsoft, and OpenAI; the signatory list has since grown past 20). Companies pledged to assess risk across the model lifecycle, publish a safety framework addressing severe risks, set explicit risk thresholds, and commit not to develop or deploy a model if mitigations can't keep its risk below those thresholds. |
| Legal status | Not legally binding — a diplomatic declaration using “we resolve” / “we affirm” language, with no enforcement mechanism. | Not legally binding on either track. The leaders' declaration is aspirational; the Frontier AI Safety Commitments are voluntary industry pledges with no enforcement mechanism. |
| Follow-up process | Closes by saying signatories “look forward to meeting again in 2024” — the commitment that became the Seoul summit. | Set the subsequent calendar: the AI Action Summit in Paris (2025), the AI Impact Summit in New Delhi (2026), and a further summit planned for Geneva (2027). |
| How it feeds into binding regulation | No direct textual link to the binding laws in this cluster. It predates California's SB 53 and New York's RAISE Act (both 2025) by roughly two years, and postdates the EU AI Act's original proposal (April 2021) by more than two years. | The voluntary company pledges anticipate obligations that later binding laws made mandatory: SB 53 and the RAISE Act both require large frontier developers to publish a safety framework addressing catastrophic risk and to report safety incidents on a fixed deadline — turning what Seoul asked 16 companies to volunteer into a statutory duty with civil penalties. The EU AI Act's GPAI/systemic-risk provisions (Articles 53 and 55) cover similar ground but were negotiated on a separate EU legislative track that began before Bletchley and became binding independently of the summit process. |
Common questions
Common questions about Bletchley Declaration (2023) vs Seoul Declaration / Frontier AI Safety Commitments (2024)
Are the Bletchley Declaration and the Seoul Declaration legally binding?
+
No. Both are diplomatic declarations, not treaties or statutes. Bletchley uses aspirational language (“we resolve,” “we affirm”) with no enforcement mechanism, and the Seoul Declaration's leaders' text is the same kind of instrument. The one Seoul-era document that reads more like a compliance framework — the Frontier AI Safety Commitments — is still a voluntary pledge by companies, not a binding rule.
Did fewer countries sign the Seoul Declaration than the Bletchley Declaration?
+
Yes. Bletchley was signed by 28 countries plus the EU. The Seoul Declaration's leaders' session was signed by 11 countries and entities: Australia, Canada, the EU, France, Germany, Italy, Japan, South Korea, Singapore, the UK, and the US. The public record doesn't explain the narrower list; the two documents were produced through different summit formats.
Do SB 53 and the RAISE Act implement the Bletchley or Seoul declarations?
+
Not formally — neither law's text cites either declaration as its legal basis. But both laws require large frontier developers to publish a safety framework and report safety incidents on a fixed deadline, which is the same category of action the Seoul Frontier AI Safety Commitments asked 16 AI companies to volunteer to do in 2024. SB 53 and the RAISE Act make mandatory, with civil penalties, what Seoul made voluntary.
What actually changed between Bletchley and Seoul?
+
Bletchley (November 2023) produced one general declaration signed by governments only, with no company-facing document. Seoul (May 2024) split into two documents: a leaders' declaration on international cooperation, and a separate Frontier AI Safety Commitments text in which 16 AI companies pledged specific practices — publishing safety frameworks, setting risk thresholds, and committing not to deploy models that cross them.
