The senators additionally requested the corporate for details about worker agreements, which may have muzzled staff who wished to alert regulators to dangers. In a July letter to the Securities and Exchange Commission, OpenAI whistleblowers stated that they had filed a criticism with the company alleging the corporate illegally issued restrictive severance, nondisclosure and worker agreements, doubtlessly penalizing staff who wished to lift considerations to federal regulators.
In an announcement to The Post earlier this month, OpenAI spokesperson Hannah Wong stated the corporate has “made important changes to our departure process to remove nondisparagement terms” from workers agreements.
The letter comes amid worker considerations that OpenAI is placing revenue earlier than security in creating its expertise. It cites a July report in The Post detailing how OpenAI rushed out its newest AI mannequin, GPT-4 Omni, to satisfy a May release date. Company leaders moved forward with the launch, regardless of worker considerations about the time-frame, and sped by means of complete security testing, undermining a July 2023 security pledge to the White House.
GET CAUGHT UP
Stories to maintain you knowledgeable
“Given OpenAI’s position as a leading AI company, it is important that the public can trust in the safety and security of its systems,” the senators wrote. “This includes the integrity of the company’s governance structure and safety testing, its employment practices, its fidelity to its public promises and mission, and its cybersecurity policies.”
“Artificial intelligence is a transformative new technology and we appreciate the importance it holds for U.S. competitiveness and national security,” OpenAI spokesperson Liz Bourgeois said in a statement. “We take our role in developing safe and secure AI very seriously and continue to work alongside policymakers to establish the appropriate safeguards going forward.”
Lawmakers, together with Sen. Chuck Grassley (R-Iowa), have stated workers at AI firms want to have the ability to provide Congress a transparent understanding of the expertise as it makes an attempt to control it — together with considerations and dangers.
Senators within the letter requested OpenAI to decide to not implementing nondisparagement agreements and “removing any other provisions” from worker agreements that could possibly be used to punish those that increase considerations about firm practices.
Senate Majority Leader Charles E. Schumer (D-N.Y.) and a bipartisan working group of senators launched suggestions earlier this yr to infuse $32 billion into AI analysis and growth, however critics have stated the plan is obscure and has stymied different efforts in Congress to craft laws. The possibilities of passing complete laws this yr are dwindling as consideration in Washington shifts to the 2024 election.
In the absence of latest legal guidelines from Congress, the White House has largely relied on voluntary commitments from the businesses to create safe and reliable AI programs. The Biden administration additionally handed a sweeping AI government order requiring firms to share testing outcomes about essentially the most highly effective fashions.
The letter additionally requested Altman whether or not OpenAI will dedicate 20 p.c of its computing assets to analysis on AI security, a dedication the corporate made final July when asserting a workforce devoted to stopping existential dangers. That group, the “Superalignment team,” has since been disbanded and its workers redistributed to different elements of the corporate.
Senate Democrats requested OpenAI if it will permit unbiased specialists to evaluate the protection and safety of its programs earlier than release, and to make its subsequent foundational AI mannequin accessible to authorities businesses for predeployment testing. Legislators additionally requested OpenAI to stipulate what misuse and security dangers its workers have noticed after releasing its most up-to-date giant language fashions.
Stephen Kohn, a lawyer representing OpenAI whistleblowers, stated Senate Democrats’ requests are “not sufficient” to treatment the chilling impact of stopping workers from talking about firm practices. “What steps are they taking to cure that cultural message,” he stated, “to make OpenAI an organization that welcomes oversight.”
Senate Democrats requested OpenAI to meet the requests by Aug. 13, together with documentation on how it plans to satisfy its voluntary pledge to the Biden administration to guard the general public from abuses of generative AI.
Kohn added that Congress should maintain hearings and an investigation into OpenAI’s practices.
“Congressional oversight on this is badly needed,” Kohn stated. “It’s essential that when you have a technology that has the potential risks of artificial intelligence that the government get in front of it.”
