OpenAI, a pioneering AI enterprise, finds itself facing an inquiry from five US senators over allegations of compromised safety protocols. The investigation, led by Brian Schatz (D-Hawaii), requests comprehensive information about OpenAI’s safety practices and its employee agreements. These allegations against OpenAI were triggered by a Washington Post report that suggested safety protocols may have been overlooked during the development of GPT-4 Omni, their latest AI model, in efforts to expedite its release.
In addition, whistleblowers from OpenAI’s now disbanded ‘superalignment team’ have also expressed worries about the restrictive nature of OpenAI’s non-disclosure agreements (NDAs).
The senators, in a stern letter to OpenAI CEO Sam Altman, asked various pointed questions, including whether the company will abide by its declaration to devote 20% of its computing resources to AI safety research, and whether it will permit independent experts to test its systems prior to release. Concerning restrictive employee agreements, the senators demanded clarification on whether OpenAI would “not enforce permanent non-disparagement agreements for current and former employees,” and if the company would commit to abolishing clauses that could be used to penalize employees who publicly express concerns about company practices.
In response to the senators’ concern, OpenAI posted reassurances of its dedication to safety, citing the company’s Preparedness Framework that is designed to counter threats posed by increasingly potent AI models. They gave assurance that they would not release a new model unless they are confident it can be introduced safely.
OpenAI also addressed the restrictive employment agreements’ concerns. They maintained that their whistleblower policy protects the right of employees to make protected disclosures and that they have amended their termination process to eliminate non-disparagement terms.
OpenAI also revealed recent initiatives to fortify its safety measures; in May, it established a new Safety and Security Committee that includes retired US Army General Paul Nakasone, a notable cybersecurity expert.
Towards passing comprehensive AI legislation in the US, the proportion is low as the attention is currently focused on the 2024 election. The White House’s response to this has been to heavily rely on AI firms’ voluntary commitments to produce safe and reliable AI systems in the absence of any new laws.
In conclusion, OpenAI’s commitment to safety and transparency is under scrutiny. Despite assurances that measures to ensure safety are in place, these allegations hint at a potential regulatory gap in the rapidly growing AI industry. Future legislation, controlling the development and deployment of AI technology, may be a solution, but it remains to be seen if this can be realised before the next election.
