Survey results show Republican voters favor required testing, third-party audits, and government authority of “high-risk” AI systems.
The American Research Institute published a survey of 750 Republican voters showing consistent support for evaluating and testing AI “high-risk” models before release.
Key Takeaways
- 77% support federal testing for national security risks before release
- 71% prefer government oversight of “high risk” models
- 85% agree AI companies should not police themselves
- 71% are in favor of requiring third-party security testing
- 85% support third-party audits of big tech developers
- 91% are in favor of safeguards to prevent economic or financial system disruption
- 81% support government authority to delay release of high-risk AI systems
83% of respondents are concerned that bad actors could exploit AI systems to harm Americans. The survey then explores how willing Republican voters are to embrace government involvement and legislation to monitor, test, and regulate AI.
Support for government safety oversight
77% of respondents think the federal government should test AI models for national security risks before release.
When dealing with “high-risk” models, 50% favor a stronger regulatory approach – creating a safety-vetting process similar to the Food and Drug Administration (FDA). 41% are more subdued, preferring formal government approval before release.
81% of Republican voters further indicated that the federal government should have the authority to delay the release of AI systems that might pose cybersecurity or biological threats.
Generally, 71% want government safety oversight, compared with just 12% who favor leaving safety in the private sector.
Independent testing over self-regulation
85% support third-party security audits, and 71% support laws requiring independent security testing. Only 17% want voluntary self-testing, and just 12% support no testing standards at all.
Further, an overwhelming 82% of respondents agree that frontier AI developers should not police themselves. This aligns with the safety concerns above.
Financial guardrails
89% of respondents agree that guardrails should be in place to prevent AI systems from disrupting the financial system or economy.
Should the US work with China?
The survey asked whether the U.S. should negotiate with China to set rules on AI use, including preventing use in cyberattacks or weapons. 77% supported such negotiations.
The survey reflects the evolving attitudes within the Republican Party since the emergence of advanced AI threat detection capabilities from Anthropic, OpenAI, and Microsoft.
