Prior Government Action on AI Security
- 9 minsSimilarly to the previous post, I compiled an introductory list of government actions on AI security while writing my master’s thesis, and I’ve decided to reproduce it here. This was originally written in February 2025, and so it is missing more recent government interventions, such as US restrictions on model releases. To bring this up to date, I intend to post more regularly on AI governance papers and government action moving forwards.
Prior Government Action on AI Security
States recognized the need for safe AI at the UK’s November 2023 Bletchley Park Summit. At the summit, 28 countries commissioned the first International AI Safety Report (published in January 2025), and stated their support for AI development that is “safe, in such a way as to be human-centric, trustworthy and responsible.”[1] Similar language has been used in 2023 and 2024 G7 documents,[2] the UN Advisory Body work on Governing AI for Humanity,[3] the UNESCO Recommendation on the Ethics of Artificial Intelligence,[4] and the OECD AI Principles.[5] However, there is little concrete detail in these documents on what “safe” and “trustworthy” mean, or on how to achieve these goals.
The May 2024 AI Seoul Summit provided more detail. States acknowledged several concrete concerns of AI researchers, including risks from chemical and biological weapons and from “manipulation and deception, or autonomous replication and adaptation conducted without explicit human approval or permission.”[6] States also recognized their “role in partnership with the private sector, civil society, academia and the international community in identifying thresholds at which the risks … would be severe without appropriate mitigations.”[7] Furthermore, they expressed intent to “promote cooperation on safety research”[8] and foster “common international scientific understanding on aspects of AI safety.”[9] However, this progress stalled at the February 2025 Paris AI Summit, with no mention of risks from AI-enabled weapons or AI action without human permission in the resulting joint statement.[10]
Unilateral government actions to address AI risk have included the US’s October 2023 Executive Order, which (although it has since been rescinded by President Trump)[11] established initial reporting requirements for companies developing large models,[12] ordered government reports on both chemical, biological, radiological, and nuclear (CBRN)[13] and cybersecurity[14] threats from AI, and committed to building standards for both developing and assessing systems,[15] calling for “robust, reliable, repeatable, and standardized evaluations.”[16] Concrete formulations of what such evaluations might look like appear in the UK’s November 2023 “Emerging Processes for Frontier AI Safety.”[17] The White House has also secured voluntary commitments from leading private organizations relating to internal and external testing, information sharing, and cybersecurity safeguards for AI model development.[18] Finally, governments have launched bodies such as the UK AI Security Institute[19] to research, test, and advise political decision-makers on the risks posed by advanced AI.[20]
Geopolitical competition over AI has included the CHIPS and Science Act passed by the US Congress in 2022, which authorized roughly $280 billion (USD) in funding to boost production of chips essential to AI. Up to $6.6 billion was awarded to Taiwan Semiconductor Manufacturing to support its development of fabrication facilities in Arizona.[21] In the same year, the US introduced export controls on semiconductors that were unambiguously designed to target China’s military AI development,[22] and it strengthened these controls in 2023.[23] The US has also attracted significant private AI investments. For example, in January 2025, the White House announced the Stargate LLC venture, which will commit up to $500 billion to the development of US AI infrastructure over 4 years.[24] In China, government venture capital funds invested roughly $209 billion in AI firms from 2013–23.[25]
As well as targeting China’s military development, the US has acted to integrate AI into its own military operations. In November 2024, Anthropic and Palantir Technologies announced that they were partnering with Amazon Web Services to provide US defence and intelligence agencies with access to Anthropic’s LLMs.[26] In addition, the US was one of several nations with an interest in lethal autonomous weapons systems (LAWS) that blocked negotiations on regulating the technology at the Sixth UN Review Conference of the Convention on Conventional Weapons (CCW).[27]
State legislation on AI has been limited to-date, but in August 2024, the world’s first AI legislation, the EU AI Act, came into force.[28] It states that developers of AI models posing “systemic risk” such as biothreats must “continuously assess and mitigate systemic risks, including for example by putting in place risk-management policies, such as accountability and governance processes, implementing post-market monitoring, taking appropriate measures along the entire model’s lifecycle and cooperating with relevant actors along the AI value chain.”[29] A code of practice to provide further guidance on complying with the EU AI Act is currently being drafted.[30] The other notable attempt at AI legislation is the State of California’s SB 1047, which would have mandated safety protocols for developers of LLMs and established liability for harms caused by models.[31] However, this was vetoed by Governor Newsom in September 2024, who argued that “By focusing only on the most expensive and large-scale models, SB 1047 establishes a regulatory framework that could give the public a false sense of security about controlling this fast-moving technology.”[32] A draft of a report into frontier AI that Newsom requested when vetoing the legislation was released in March 2025.[33]
Footnotes
[1]: ‘The Bletchley Declaration by Countries Attending the AI Safety Summit, 1–2 November 2023’ (Bletchley Park: AI Safety Summit, 1 November 2023), https://www.gov.uk/government/publications/ai-safety-summit-2023-the-bletchley-declaration/the-bletchley-declaration-by-countries-attending-the-ai-safety-summit-1-2-november-2023.
[2]: ‘Hiroshima Process International Guiding Principles for Organizations Developing Advanced AI Systems’ (G7, 30 October 2023); ‘Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems’ (G7, 30 October 2023); ‘G7 Science and Technology Ministers’ Meeting Communiqué’ (G7, 19 July 2024).
[3]: ‘Governing AI for Humanity’ (New York: UN Advisory Body on Artificial Intelligence, September 2024), https://digitallibrary.un.org/record/4062495.
[4]: ‘Recommendation on the Ethics of Artificial Intelligence’ (UNESCO, 23 November 2021), https://unesdoc.unesco.org/ark:/48223/pf0000381137.
[5]: ‘AI Principles’ (OECD, 2024), https://www.oecd.org/en/topics/ai-principles.html.
[6]: ‘Seoul Ministerial Statement for Advancing AI Safety, Innovation and Inclusivity’.
[7]: Ibid.
[8]: ‘Seoul Declaration for Safe, Innovative and Inclusive AI by Participants Attending the Leaders’ Session: AI Seoul Summit, 21 May 2024’ (Seoul: AI Seoul Summit, 21 May 2024), https://www.gov.uk/government/publications/seoul-declaration-for-safe-innovative-and-inclusive-ai-ai-seoul-summit-2024/seoul-declaration-for-safe-innovative-and-inclusive-ai-by-participants-attending-the-leaders-session-ai-seoul-summit-21-may-2024.
[9]: ‘Seoul Statement of Intent toward International Cooperation on AI Safety Science, AI Seoul Summit 2024 (Annex)’ (Seoul: AI Seoul Summit, 21 May 2024), https://www.gov.uk/government/publications/seoul-declaration-for-safe-innovative-and-inclusive-ai-ai-seoul-summit-2024/seoul-statement-of-intent-toward-international-cooperation-on-ai-safety-science-ai-seoul-summit-2024-annex.
[10]: ‘Statement on Inclusive and Sustainable Artificial Intelligence for People and the Planet.’ (Paris, France: Artificial Intelligence Action Summit, 11 February 2025), https://www.elysee.fr/en/emmanuel-macron/2025/02/11/statement-on-inclusive-and-sustainable-artificial-intelligence-for-people-and-the-planet.
[11]: ‘Removing Barriers to American Leadership in Artificial Intelligence’.
[12]: ‘Executive Order 14110: Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence’ (Washington, D.C.: President of the United States, 30 October 2023), sec. 4.2, https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/.
[13]: Ibid., sec. 4.4(a)(i).
[14]: Ibid., sec. 4.3(b)(iii).
[15]: Ibid., sec. 4.1.
[16]: Ibid., sec. 2(a).
[17]: ‘Emerging Processes for Frontier AI Safety’ (AI Safety Summit, October 2023), https://www.gov.uk/government/publications/emerging-processes-for-frontier-ai-safety.
[18]: ‘Biden-Harris Administration Secures Voluntary Commitments from Eight Additional Artificial Intelligence Companies to Manage the Risks Posed by AI’ (Washington, D.C.: President of the United States, 12 September 2023), https://bidenwhitehouse.archives.gov/briefing-room/statements-releases/2023/09/12/fact-sheet-biden-harris-administration-secures-voluntary-commitments-from-eight-additional-artificial-intelligence-companies-to-manage-the-risks-posed-by-ai/.
[19]: ‘Introducing the AI Safety Institute’ (UK AI Safety Institute, 17 January 2024), https://www.gov.uk/government/publications/ai-safety-institute-overview/introducing-the-ai-safety-institute.
[20]: For information on other government institutes, see Renan Araujo, ‘Understanding the First Wave of AI Safety Institutes: Characteristics, Functions, and Challenges’ (Institute for AI Policy and Strategy, 7 October 2024), https://www.iaps.ai/research/understanding-aisis.
[21]: ‘Biden-Harris Administration Announces CHIPS Incentives Award with TSMC Arizona to Secure U.S. Leadership in Advanced Semiconductor Technology’, U.S. Department of Commerce, 15 November 2024, https://www.commerce.gov/news/press-releases/2024/11/biden-harris-administration-announces-chips-incentives-award-tsmc.
[22]: ‘Export Controls on Semiconductor Manufacturing Items’ (Washington, D.C.: Department of Commerce, Bureau of Industry and Security, 25 October 2023), https://www.federalregister.gov/documents/2023/10/25/2023-23049/export-controls-on-semiconductor-manufacturing-items.
[23]: ‘Implementation of Additional Export Controls: Certain Advanced Computing Items; Supercomputer and Semiconductor End Use; Updates and Corrections; and Export Controls on Semiconductor Manufacturing Items; Corrections and Clarifications’ (Washington, D.C.: Department of Commerce, Bureau of Industry and Security, 4 April 2024), https://www.federalregister.gov/documents/2024/04/04/2024-07004/implementation-of-additional-export-controls-certain-advanced-computing-items-supercomputer-and.
[24]: Sareen Habeshian, ‘Trump Announces Billions in Private Sector AI Investment’, Axios, 21 January 2025, https://www.axios.com/2025/01/21/trump-announces-billions-in-private-sector-ai-investment.
[25]: Martin Beraja et al., ‘Government as Venture Capitalists in AI’, Working Paper, Working Paper Series (National Bureau of Economic Research, July 2024), 2, https://doi.org/10.3386/w32701.
[26]: ‘Anthropic and Palantir Partner to Bring Claude AI Models to AWS for U.S. Government Intelligence and Defense Operations’, Business Wire, 7 November 2024, https://www.businesswire.com/news/home/20241107699415/en/Anthropic-and-Palantir-Partner-to-Bring-Claude-AI-Models-to-AWS-for-U.S.-Government-Intelligence-and-Defense-Operations.
[27]: ‘Killer Robots: Military Powers Stymie Ban’ (Human Rights Watch, 19 December 2021), https://www.hrw.org/news/2021/12/19/killer-robots-military-powers-stymie-ban.
[28]: ‘Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 Laying down Harmonised Rules on Artificial Intelligence and Amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828’, OJ L, 2024/1689, 12.7.2024 § (2024), https://eur-lex.europa.eu/eli/reg/2024/1689/oj.
[29]: Ibid., sec. 115.
[30]: ‘Third Draft of the General-Purpose AI Code of Practice Published, Written by Independent Experts’ (European Commission, 11 March 2025), https://digital-strategy.ec.europa.eu/en/library/third-draft-general-purpose-ai-code-practice-published-written-independent-experts.
[31]: Scott Wiener et al., ‘Safe and Secure Innovation for Frontier Artificial Intelligence Models Act.’, Pub. L. No. SB 1047 (2024), https://legiscan.com/CA/text/SB1047/id/3019694.
[32]: Gavin Newsom, ‘Veto Message: SB 1047’ (Office of the Governor, State of California, 29 September 2024).
[33]: Jennifer Tour Chayes, Mariano-Florentino Cuéllar, and Li Fei-Fei, ‘Draft Report of the Joint California Policy Working Group on AI Frontier Models’ (Joint California Policy Working Group on AI Frontier Models, 18 March 2025), https://www.cafrontieraigov.org/.
Alex Chalk
AI Research | Ontario Graduate Scholar | Software Developer