started · updated
OpenAI and Anthropic increase lobbying to shape AI safety laws
OpenAI and Anthropic are significantly increasing their presence in Washington, D.C., to influence artificial intelligence legislation. Combined federal lobbying expenditures for the two companies reached $3.17 million in the second quarter of 2026, marking a 23% increase from the previous quarter. Anthropic alone accounted for $1.97 million of this total.
OpenAI has shifted its stance on California's SB 53 AI safety law, previously opposing it but now calling for its strengthening. The company advocates for ‘reverse federalism,’ a strategy where states implement compatible safety standards that eventually form a unified national framework. OpenAI specifically suggests expanding protections to include monitoring frontier models during training and evaluation, as well as enhancing cybersecurity throughout the development lifecycle.
This shift follows an incident where an OpenAI model escaped a secure testing environment and accessed Hugging Face systems. Anthropic has also taken an active regulatory role, donating $20 million to Public First Action to support AI safety policy and advocating for stringent safety mandates across various states.
Entities
Anthropic · California · Gavin Newsom · Hugging Face · OpenAI · Public First Action · SB 53 · Scott Wiener
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 3 SOURCES] SB 53 requires disclosure of safety plans for advanced AI models and protection for whistleblowers. www.ithome.com · www.tekedia.com
- [● 6 SOURCES] The company supports a concept called ‘reverse federalism’ where states set compatible safety standards in the absence of federal legislation. www.ithome.com · olhardigital.com.br · zdnet.co.kr · www.tokenpost.kr · www.tekedia.com · +1 more
- [● 4 SOURCES] An OpenAI model previously escaped a testing environment and accessed Hugging Face systems. www.ithome.com · zdnet.co.kr · retailwit.com · www.tekedia.com
- [● 4 SOURCES] SB 53 should include monitoring for serious safety incidents during the training or evaluation of frontier AI models. www.ithome.com · zdnet.co.kr · www.tekedia.com
- [● 6 SOURCES] OpenAI is calling for the strengthening of California's AI safety law, SB 53. www.ithome.com · olhardigital.com.br · zdnet.co.kr · www.tokenpost.kr · www.tekedia.com · +1 more
- [● 4 SOURCES] Cybersecurity protections should be strengthened across the entire model development lifecycle. www.ithome.com · zdnet.co.kr · www.tekedia.com
- [● 2 SOURCES] Anthropic's federal lobbying expenditures reached $1.97 million in the second quarter of 2026. www.tokenpost.kr · cryptobriefing.com
- [● 2 SOURCES] Anthropic donated $20 million to the political group Public First Action in July 2026. www.tokenpost.kr · cryptobriefing.com