Republican Senator Chuck Grassley demanded OpenAI turn over documents proving it does not silence employees who wish to share concerns with federal regulators. The letter, obtained exclusively by The Washington Post, highlights the growing bipartisan push to ensure OpenAI's AI is developed safely.
Grassley asked OpenAI CEO Sam Altman to outline changes to employment agreements to protect employees who wish to voice concerns without penalty. According to The Post, the request comes after employees warned about rushed safety testing of the GPT-4 Omni model.
Grassley's letter follows another sent by five senators, led by Brian Schatz, requesting information on how OpenAI will meet its public safety commitments. The message comes amid concerns that OpenAI prioritizes profits over security, accelerating security testing to meet release dates. Grassley also alludes to a previous controversy over whether OpenAI silences employees through restrictive covenants. In July, The Washington Post reported that whistleblowers filed complaints alleging severance and confidentiality agreements that would penalize those who disclosed concerns to regulators.
Hannah Wong, a spokeswoman for OpenAI, said in July that the company eliminated restrictive terms in exit agreements. Liz Bourgeois, another spokeswoman, reiterated that OpenAI's policy protects employees' rights to raise issues with any government agency.
The rapid advancement of AI has sharpened policymakers' concerns about the technology industry's power. In the United States, AI companies operate largely unregulated, and policymakers are seeking the help of whistleblowers to formulate new policies, according to Infobae. Grassley stressed the importance of employees being able to disclose protected concerns without restrictions. Clare Slattery, Grassley's spokeswoman, said the senator's staff requested documents on employment agreements from OpenAI, but they still need to be provided.
In a recent statement on X, Altman said OpenAI wants employees to feel comfortable raising concerns. In May, it rescinded non-disparagement terms. OpenAI also responded to the senators by saying it consulted with more than 100 experts to assess the risks of GPT-4 Omni.
Finally, Grassley asked OpenAI to detail how many times employees have asked to raise concerns with federal authorities and whether the company is under investigation by the Securities and Exchange Commission. The company has until August 15 to respond.
