Questions of trust and safety loomed over the event. The day before the new products launched, OpenAI apologized for its agents’ hacking of an Australian government website. “We are sorry and are working to do better in the future,” he said in a blog post.
The incident, which occurred in June, involved an internal model under development that accessed non-public information related to Australia’s public health insurance program. On Tuesday, Altman and OpenAI executives stressed that the episode — and other major incidents of OpenAI systems going malicious or acting unintentionally — all involved models that were not intended for public release and were distinct from commercial offerings available to users.
“The instances we disclosed were running internal-only models that were not close to shipping,” said Alex Embericos, product manager at OpenAI.
Despite this distinction, Altman said there is a tradeoff between pushing the boundaries of AI development and building cutting-edge AI systems safely.
“To push the models to a new level of capability, it will take us some time to figure out how to ensure that alignment, surveillance, safety and security stay well ahead of capabilities,” he told reporters after his opening remarks. “But you can see that we were able to create a huge amount of useful new stuff without a major new model, and there will be new models.”
On Monday, OpenAI announced that it would refrain from releasing its latest model, GPT-6.1 Astra, due to concerns that the system would not follow human instructions. “While (GPT-6.1 Astra) has improved on areas such as model laziness, it hasn’t quite hit the mark in terms of meeting scope and clearance,” security systems manager Saachi Jain said in a statement.
OpenAI executives also acknowledged that its new products would require significant computing resources, with Altman classifying Dots as a “premium product.” He said the company hopes to eventually roll out a version of Dots to a mainstream audience in the future, citing computational intensity as the reason for the soft launch.
Points are only available to users subscribed to OpenAI’s Pro plan, which costs between $100 and $500 per month, as well as Business Premium and select Enterprise users. Points are also not available in Europe and the United Kingdom, which Altman attributed to the region’s strict regulations.
“We have a great track record of launching our products in the EU, but they have a process that takes longer, and sometimes unpredictably,” Altman told NBC News. “But we really want to ship all our goods to Europeans.”
Altman revealed Dots a few weeks after the creator of Facebook and Instagram Meta released his AI agent series Muse, which he bills as “your personal agent.” The Muse app reached the top spot on the Apple and Android app stores, quickly surpassing Elon Musk’s xAI agentic AI offering Grok Bot.
OpenAI also unveiled a new model, GPT-6.1 Sol, which it says almost matches its top-of-the-line Astra model in terms of performance for a fraction of the price, in addition to announcing strengthened privacy protocols and a new collaboration space for users to interact with its systems – aptly called “ChatGPT Space”.
While Altman and OpenAI employees promoted their new offerings, OpenAI President Greg Brockman was meeting with President Donald Trump and dozens of other technology leaders across the country to lay out an ambitious vision for AI. Trump used the opportunity to “usher in the era of superintelligence.”
Altman said he hoped the Washington meeting would allow AI titans to build consensus on AI risks while allowing society to harness the technology’s potential benefits.
Altman said he disagreed with industry executives who refused to acknowledge that AI could pose significant risks — “you can imagine a legitimate loss of control of an AI because we don’t take safety and security seriously enough” — while distancing themselves from claims that AI could kill all of humanity within a few years.
“I hope in this industry we can come together and say we need to take that middle path between the two,” he said.
Gn bussni
