OpenAI’s release of GPT-5.6 and its integration of the model into ChatGPT Work, following a U.S. government review, underscores the changing regulatory landscape for AI models. It also reflects how vendors are trying to demonstrate to enterprises how their models can power work across their organizations.
OpenAI officially launched the GPT-5.6 family of models on July 9, two weeks after previewing GPT-5.6 Sol. The models consist of Sol, the vendor’s new flagship model; Terra, a model for enterprise work; and Luna, OpenAI’s least expensive model. OpenAI said Sol is adept at coding, knowledge work, cybersecurity and science.
GPT-5.6 arrived in full release a day after OpenAI competitor SpaceXAI, introduced its Grok 4.5 model, emphasizing its significant coding capabilities. But OpenAI’s latest model lineup also shows how the ChatGPT creator continues to move away from its consumer-based roots.
“The company itself is maturing as an enterprise solution provider,” said Lian Jye-Su, an analyst at Omdia, a division of Informa TechTarget. He added that GPT-5.6 is significant because OpenAI has improved its agentic reasoning and coding capabilities, indicating that the vendor continues to strengthen its enterprise offerings.
Government Evaluation
The model family also arrives after a two-week review by the U.S. government, which assessed the model for national security risks because the Sol model exceeded the cybersecurity risk threshold. The government decided to review GPT-5.6 after rival Anthropic had to shut down its Fable 5 and Mythos 5 models initially, and after the U.S. imposed export controls on them. Anthropic later turned the model back on after the government gave a go-ahead.
With GPT-5.6 being the second AI model to require government review before its full release to the public, the government sees the models as more than technical advancements.
“We are looking at a world where AI models are being viewed as critical national assets,” Su said. Among the Trump administration’s concerns are foreign AI vendors apparently using extensive model distillation on U.S. companies’ models.
“It’s about trying to make sure that whatever they release to the market that has an opportunity to come in touch with foreign model developers, critical data won’t get leaked or get distilled by other entities that may have malicious intent,” Su said. He added that testing the model reassures the U.S. government that critical information is not being leaked.
However, while the situation with Fable 5, Mythos 5, and now GPT-5.6 has changed the risk assessment of AI models in the U.S., the lack of a consensus might feel unsettling, said Nick Patience, an analyst at Futurum Group.
“The U.S. regulatory environment for AI models at the moment is a bit chaotic,” Patience said. He added that it is clear the administration wants to be involved in how the model is released. “The fact that it’s now not only Anthropic that’s getting treated like this … gives you some feeling that on a forward basis there’s also going to be some sort of evaluation of each model before it comes out.”
GPT-5.6 and ChatGPT Work
Other than the change in the regulatory environment, the release of GPT-5.6 seems more like an iterative design than an overhaul, Su said.
“It doesn’t quite have, let’s say, the aura that’s generated by Mythos 5,” he said. “It’s not, let’s say, earth-shattering, but then it’s still moving in the right direction.”
Also, OpenAI’s enterprise focus is even more evident in its release of ChatGPT Work. ChatGPT Work is an AI agent that does more than answer questions. It connects to applications such as Slack, Google Drive, Gmail, Jira and Salesforce. It combines OpenAI’s Codex agent into a unified desktop app for writing, debugging and running code locally. It builds and hosts internal portals and live dashboards, and it can also generate spreadsheets, slides, reports and documents. GPT-5.6 powers ChatGPT Work.
OpenAI is catching on to what Anthropic realized when it introduced Claude Cowork. The models are the engines that need to power the work being done in the enterprise, Patience said.
“ChatGPT Work is OpenAI betting that enterprises will pay for some sort of finished output, not just a text answer,” he said. “It’s stopping the idea that ChatGPT is just a conversation to be had or a tool to have a conversation with and start being some sort of co-worker that gives you deliverables.”
He added that the choice to make ChatGPT Work exclusively cloud-based might create challenges due to potential data residency and privacy requirements, which could mean some regulated verticals might have to consider how to implement it without violating those requirements.
