Every major AI company has a "responsible AI" page. They publish principles. They sign pledges. They testify before Congress about the importance of safety. Then they hire lobbyists.
What happens when you compare what AI companies say about regulation with what they actually lobby for? We pulled public statements, lobbying disclosures, PAC filings, and legislative testimony to build a side-by-side comparison. The gaps range from subtle to staggering.
The Methodology
For each company, we examined:
- Public statements: Official blog posts, CEO speeches, congressional testimony, and published AI principles
- Lobbying disclosures: Federal and state lobbying filings identifying specific bills and issues
- PAC contributions: Campaign donations to candidates with known AI policy positions
- Legislative positions: Documented support or opposition to specific AI bills
Here's what we found, company by company.
OpenAI
What they say: "We believe AI should be developed safely and in a way that benefits humanity." CEO Sam Altman has called for federal AI regulation, testified before Congress advocating for a licensing regime, and published detailed safety frameworks.
What they lobby for: OpenAI has lobbied against multiple state AI safety bills, including California's SB 1047 (which would have required safety testing for large AI models). Their federal lobbying focuses on preempting state regulation โ effectively arguing that only Congress should regulate AI, while Congress passes nothing. They've also lobbied against copyright liability for AI training data and against mandatory disclosure of training data sources.
The gap: OpenAI wants regulation โ but only the kind they help write, at the federal level, where the legislative process is slow enough to never actually produce binding rules. State laws that might impose immediate requirements? Those they fight.
Meta
What they say: "Open source AI is safer because it allows the entire community to inspect and improve models." Meta positions its open-source Llama models as a safety advantage and publishes extensive responsible AI research.
What they lobby for: Meta spent $36M+ in H1 2025 alone on lobbying, making it the largest single AI lobbying spender. Their top priorities include federal preemption of state AI laws, blocking requirements for safety testing of open-source models, preventing regulation of AI-generated content on social platforms, and securing favorable data center permitting rules. They've actively lobbied against bills that would require disclosure of AI-generated content in political advertising.
The gap: Meta argues open source is inherently safer โ while lobbying to ensure no one can require them to prove it's safer through mandatory testing or auditing. Their "openness" argument is simultaneously a safety claim and a shield against regulation.
Anthropic
What they say: "We believe in thoughtful AI regulation." Anthropic has the most explicitly pro-regulation public posture of any major AI lab. CEO Dario Amodei has published detailed proposals for AI governance. The company's Responsible Scaling Policy commits to safety evaluations at each capability threshold.
What they lobby for: Anthropic contributed $20 million to Public First Action, a PAC that backs candidates who largely oppose binding AI regulation. Their lobbying disclosures show engagement on AI safety standards โ but primarily in the context of voluntary frameworks rather than mandatory requirements. They've supported federal preemption language that would override stronger state-level rules.
The gap: Anthropic talks the best safety talk in the industry. But their $20M PAC contribution went to back candidates who oppose the kind of mandatory regulation Anthropic publicly endorses. If you believe in regulation, funding anti-regulation candidates is a strange way to show it.
What they say: Google published its AI Principles in 2018 โ one of the first major companies to do so. The principles include commitments to avoid AI that causes harm, to be socially beneficial, and to incorporate privacy design principles. Google regularly publishes AI safety research and participates in industry safety initiatives.
What they lobby for: Google is a top-5 overall federal lobbying spender, and AI has become an increasing share of that spend. Their lobbying targets include: weakening AI bias auditing requirements, opposing mandatory algorithmic transparency, securing favorable AI procurement rules for government contracts, and blocking state-level AI regulation. Google's lobbying operation spans over 100 individual lobbyists and dozens of firms.
The gap: Google's AI Principles commit to "avoiding creating or reinforcing unfair bias" โ while their lobbyists work to weaken the bias auditing requirements that would verify whether they're keeping that commitment. The principles are voluntary; they're lobbying to make sure they stay that way.
Microsoft
What they say: Microsoft's "AI for Good" program commits $40M+ annually to AI projects addressing societal challenges. President Brad Smith has been one of the most vocal tech executives calling for AI regulation, including proposals for AI licensing, safety requirements, and international cooperation.
What they lobby for: Microsoft's lobbying operation โ one of the largest in tech โ focuses on securing its competitive position through its OpenAI partnership. They've lobbied for AI procurement rules that favor established vendors (like Microsoft Azure), against regulations that would require disclosure of AI model capabilities or training data, and for federal preemption of state AI laws. Their lobbying also targets antitrust concerns about the OpenAI investment.
The gap: Microsoft calls for AI regulation publicly while lobbying for rules that would cement its competitive advantages and shield its OpenAI partnership from scrutiny. "AI for Good" runs on $40M/year; Microsoft's overall lobbying budget dwarfs that figure. The question is which investment more accurately reflects the company's priorities.
The Pattern
Across all five companies, the same pattern repeats:
- Public commitment to safety and responsibility โ genuine or not, every company feels pressure to say the right things
- Lobbying against mandatory enforcement โ voluntary commitments are fine; legal requirements are fought
- Support for federal preemption โ every company wants Congress (which moves slowly) to override states (which move fast)
- Political spending misaligned with public positions โ PAC money flows to candidates who oppose the regulation companies publicly endorse
This doesn't necessarily mean the companies are acting in bad faith. Corporate lobbying is complex, and companies can simultaneously believe in responsible AI and believe specific proposed regulations are poorly designed. But when the pattern is universal โ when every company supports regulation in principle and opposes it in practice โ the reasonable conclusion is that the public statements are primarily reputation management.
Why This Tracker Matters
Voters, legislators, and journalists shouldn't have to take AI companies at their word. When Anthropic says it supports regulation, we can check their PAC spending. When OpenAI says safety comes first, we can check their lobbying disclosures. When Google invokes its AI Principles, we can check which bills they opposed.
That's what this site does. Browse every company's full profile โ public statements, lobbying data, PAC contributions, and legislative positions โ all in one place.