In an X post Monday commenting on DeepSeek’s sudden success, OpenAI CEO Sam Altman promised to “pull up some releases” and it appears he has done so. OpenAI unveiled its newest product on Tuesday, a “tailored version of ChatGPT designed to provide U.S. government agencies with an additional way to access OpenAI’s frontier models,” per the announcement post. ChatGPT Gov will reportedly offer even tighter data security measures than ChatGPT Enterprise, but how will it handle the hallucinations that plague the company’s other models?
According to OpenAI, more than 90,000 federal, state, and local government employees across 3,500 agencies have queried ChatGPT more than 18 million times since the start of 2024. The new platform will enable government agencies to enter “non-public, sensitive information” into ChatGPT while it runs within their secure hosting environments — specifically, the Microsoft Azure commercial cloud or Azure Government community cloud — and cybersecurity frameworks like IL5 or CJIS. This enables each agency to “manage their own security, privacy and compliance requirements,” Felipe Millon, Government Sales lead at OpenAI told reporters on the press call Tuesday.
deepseek’s r1 is an impressive model, particularly around what they’re able to deliver for the price.
we will obviously deliver much better models and also it’s legit invigorating to have a new competitor! we will pull up some releases.
— Sam Altman (@sama) January 28, 2025
The company also provided a demonstration of what ChatGPT Gov, which runs atop the GPT-4o model, can do during the press call. In the demo, a “federal employee” signed into ChatGPT, generated a five-week plan for their job duties and then analyzed a printed version of the same 5-week plan marked up with hand-labeled notes and addenda. ChatGPT then drafted a memo to the agency’s legal and compliance department summarizing its plan.
“I know President Trump is also looking at how we can potentially streamline that, because it’s one way of getting more modern software tooling into the government and helping the government run more efficiently,” OpenAI CPO Kevin Weil told CNBC. “So we’re very excited about that.”
ChatGPT Gov is built largely upon the existing ChatGPT Enterprise framework, which is still going through the Federal Risk and Authorization Management Program and has not yet been cleared to handle non-public data. There is no timeline yet for when that accreditation might come through but the AI likely won’t be rolled out to actual agency employees before that happens.
Even with the focus on data and operational security, relying on generative AI for such vital and sensitive tasks as the functioning of government poses unique and difficult challenges. We’ve already seen more than one lawyer get caught using chatbots to “perform research,” resulting in them citing hallucinated case law, as well as police rely on faulty AI-based evidence to wrongfully arrest people. California voters also recently rejected a measure that would have put AI in charge of bail sentencing.
What’s more, a 2024 Stanford HAI study found that popular legal models from Thomas Reuters and LexisNexis, which promise “hallucination-free legal research,” actually did hallucinate in their responses, 34% and 17% of the time, respectively. The study’s authors point out that’s still a significant improvement over general-purpose chatbots which hallucinated responses to legal queries anywhere between 58% and 82% of the time.
OpenAI’s announcement came just hours after DeepSeek revealed V3, its latest generative model, which offers comparable performance to OpenAI’s state of the art o1 reasoning model while using 20 to 50 times less power, depending on the task. The Chinese startup has thrown the American AI industry into disarray with that revelation, calling into question the need to continually build more and more data centers and power plants if more efficient AI code could accomplish the same. Nvidia, the leading global supplier of GPUs, the specialized processing units that AI systems require for training and inference tasks, lost $600 billion in market capitalization in response to the DeepSeek news. Gas and nuclear energy companies saw their stock values decline as well.
