By Tom LeithauserOpenAI has offered inadequate responses to lawmakers’ questions about the company’s AI (artificial intelligence) models’ misbehavior, according to 20 House Democrats who wrote yesterday to OpenAI Chief Executive Officer Sam Altman.The letter noted that OpenAI reportedly fired three “whistleblowers” over their outreach to an AI security researcher and has offered incomplete and contradictory responses to previous congressional queries about the company’s AI agents hacking Hugging Face, Inc., and other organizations (CPR, Sept. 3).“OpenAI has previously shown a disregard for safety concerns highlighted by your employees. Months before the Hugging Face incident, two OpenAI employees warned executives that your newest models were not being appropriately monitored during testing, and executives responded that the tests needed to move forward as quickly as possible to release the models on time,” the lawmakers told Mr. Altman.“These firings also follow a slew of new cybersecurity incidents involving OpenAI models. OpenAI and Anthropic are reportedly investigating tens of thousands of incidents, most of them involving OpenAI models. Your company has shared varying degrees of information about these incidents,” they said.“You have said that your company has paused training, testing, and tool use for its most capable models and would only resume training ‘when we are confident that we have additional safeguards and alignment improvements in place,’” they recalled.“These incidents appear to stem from alignment issues with your models. Your own misalignment reports describe models writing instructions into their own summaries to hide failures and to tell later copies to ignore developer messages. Until these issues are resolved, it is not clear that these models can be developed safely,” they warned.“We demand that you commit to not firing whistleblowers sharing information related to the safety of your models and that you keep your current pause on your most capable models in place until independent third-party oversight is in place,” they said. They also asked Mr. Altman to respond by Nov. 13 to a lengthy list of questions about the firings and security incidents involving OpenAI’s models.The letter’s lead signatory was Rep. Greg Casar (D., Texas). OpenAI did not respond to a request for comment.MainStory: TopStory FederalLegislation DataSecurity AINews
Click Here For The Original Source.
——————————————————–
