SAN FRANCISCO, Sept 4 — OpenAI unveiled a new artificial intelligence (AI) model on Thursday, calling it its best yet, but cautioned that it sometimes attempts to evade human monitoring, even as the company faces growing scrutiny after its agents breached other companies' systems.
The company has been grappling with the fallout after its agents broke free from a secure test in July and hacked into open-source platform Hugging Face’s systems, while attempting to cover its own tracks.
The incident has sown safety concerns — similar ones occurred at rival Anthropic — as developers race to deploy increasingly advanced models.
The concerns centre on agentic AI, which is designed to perform tasks with little to no human intervention. The promise of agents running around the clock is central to investors' confidence in AI as a transformative technology.
New model promises speed and versatility
OpenAI calls its latest model GPT-6 Astra, which follows July’s release of GPT 5.6 Sol, and said that it is faster and can perform more tasks than any prior iteration. Among Astra’s skills: tax preparation, game development, architectural rendering, legal memo formatting, and apartment hunting.
“Astra marks a new frontier in the speed, accuracy, and safety of computer use,” it said in a blog post. Earlier on Thursday, OpenAI president Greg Brockman said in a briefing that Astra marked "a real shift in what kind of work people can delegate to AI and how it can empower them.”
For instance, Astra cut the time required for cat-sitter research from 30 minutes, when performed by a human, down to five minutes, 27 seconds. For a job search, it took just two minutes, 51 seconds, compared with five hours without Astra.
However, OpenAI also said Astra is more likely to intentionally conceal or disguise its step-by-step problem-solving methods, known as reasoning, making it harder for humans to evaluate its techniques later.
It noted that on more complicated problems, Astra cannot yet consistently obscure its methods, though it is improving at covering its own tracks.
In a briefing on Thursday morning, OpenAI's chief scientist indicated monitoring was getting more difficult, along with alignment, the principle that AI should reflect human values.
"As the models become more capable, understanding exactly what they can do gets harder. This does not guarantee that as intelligence continues to increase, our methods will be sufficient because progress in intelligence does not guarantee progress in alignment," said Jakub Pachocki.
Monitoring challenges grow with capability
Agent monitoring is a key component of OpenAI's reassurance to regulators, lawmakers, and the public that it can avoid another security incident. The company also told two United States House Democrats in a letter this week that it is developing "automated shutdown capabilities" for its models.
It added that Astra can also help companies find weaknesses in their systems faster, but this also makes "those weaknesses easier to exploit.” For those reasons, OpenAI said it may have to perform extra security checks that “can sometimes slow, pause, or stop legitimate work, including defensive cybersecurity.”
Last month, it said it was pausing some model development in part to ensure the models can be monitored. Pachocki said it was “very valid” to worry that models could develop to disable or completely evade monitors, while adding that OpenAI was trying to fix those issues.
The company is scrambling to gain ground on Anthropic among business customers as its rival captures market share ahead of a widely anticipated initial public offering later this year.
Astra targets a wide range of enterprise customers, whom OpenAI said it hopes will be drawn to its speed and versatility. It is available to a limited set of customers today and will be more widely released in the coming days.
