The AI race is hitting the brakes. The question is for how long

OpenAI’s decision to shelve GPT-6.1 Astra adds to signs that frontier AI developers are becoming more willing to slow or restrict releases when safety systems fail to keep pace.


The race to build ever more powerful artificial intelligence systems is showing its clearest signs yet of hitting the brakes.

OpenAI this week shelved the planned October release of GPT-6.1 Astra after the model failed to meet internal safety standards, according to Reuters. The decision does not amount to an industrywide halt, and OpenAI is still releasing other models and agent products. But it adds to a growing pattern in which frontier AI companies are slowing, restricting or reconsidering systems whose capabilities appear to be advancing faster than their safeguards.

The immediate problem is no longer simply whether a model will produce an offensive or dangerous answer in a chat window. Increasingly powerful systems can browse websites, write and execute code, call outside tools and pursue a task across many steps with limited human supervision. That makes failures harder to predict and potentially far more consequential.

The shift is harder to dismiss after a series of agent incidents. In July, an OpenAI agent escaped a controlled testing environment and broke into systems at Hugging Face while trying to complete an evaluation, Reuters reported. OpenAI later said it was reinforcing its safeguards.

For an industry accustomed to treating every major capability jump as something that should quickly become a product, shelving a flagship model is notable.

“The newsworthy event is not a lab that pulls a model,” Kuber Sharma, senior director of product marketing at UiPath, told AI Safety Watch in an email interview. “It is a lab that ships a model that should have been pulled and discovers this six months post-deployment, when the surface area of the failure is much larger.”

A different race

The apparent slowdown has been building rather than arriving all at once.

OpenAI has said it paused training its most capable models until additional safeguards are in place. Anthropic CEO Dario Amodei has separately called for the industry to slow the pace at the frontier. Current and former researchers from major AI labs have also warned that competition can push development ahead faster than safety practices mature, Reuters reported this week.

Yet “slowdown” remains a relative term.

OpenAI is still shipping products. At its developer event this week, the company introduced its dots agents and GPT-6.1 Sol, underscoring that it has not stepped away from the frontier race. The emerging distinction is narrower: developing more capable AI is continuing, but releasing every capability gain at full scale is no longer automatic. Reuters reported that dots are designed to act across applications under new controls.

Arjun Jaggi, an applied AI researcher and executive adviser, told AI Safety Watch in an email interview that withholding a model suggests a lab’s evaluation process has detected capabilities or behavior it cannot adequately contain with its existing safeguards.

“A lab holding back a model is a signal worth taking seriously,” Jaggi said. “But it’s not a substitute for independent verification of why.”

The larger problem, he said, is that the same companies building frontier systems generally determine whether those systems are safe enough to deploy.

“The public and regulators currently have to take the lab’s word for both the risk and the adequacy of the decision,” Jaggi said.

From answers to actions

The growing emphasis on AI agents makes that judgment more consequential.

For most of the chatbot era, a model primarily returned information to a person. A bad answer could still cause harm, but the system itself generally could not alter a database, send an email, deploy software or move through a company’s internal systems on its own.

Agents change that equation.

“The risk is not what the model knows. It is what the model can do,” Sharma said. “An autonomous agent that reads email, writes code, executes workflows and triggers deployments has a larger blast radius than any previous enterprise tool.”

That concern grows as agents receive persistent credentials and access to multiple systems.

Noe Ramos, vice president of AI operations at Agiloft, told AI Safety Watch in an email interview that moving from generating information to acting on it creates a fundamentally different security problem.

“A model that drafts text carries one kind of exposure,” Ramos said. “An agent that writes to systems, executes workflows and triggers downstream processes without human review carries a fundamentally different one.”

Among the risks, Ramos said, are privilege escalation, prompt injection through outside information and chains of agents producing outcomes that become difficult to attribute or reverse.

Prompt injection is particularly troublesome because an attacker may not need to compromise the AI system itself. Instructions hidden in an email, document or webpage can potentially manipulate an agent that reads the material and has permission to take actions.

Nishanth Sirikonda, a cloud enterprise architect at FirstDay Foundation, told AI Safety Watch in an email interview that an agent encountering a compromised document could expose credentials, transfer sensitive information or make unauthorized changes before a person realizes anything has gone wrong.

“A model that refuses a harmful request in chat may still be unsafe when it has credentials, tools and time to act,” Sirikonda said.

That is one reason conventional model evaluations may be insufficient. Testing whether a chatbot refuses a prohibited request says relatively little about what happens when the same underlying system is given a browser, authentication credentials and hours to accomplish a loosely defined objective.

The same problem appears in different forms across the responses AI Safety Watch received. Eshaan Jain, a senior product manager at Mphasis supporting T-Mobile, told AI Safety Watch in an email interview that persistent access across multiple systems can leave agents with privileges long after a task ends. Afsheen Afshar, founder and managing partner of Pilot Wave Holdings, said small gaps in instructions can compound over long chains of action. “A small gap in the instructions can compound over a long chain of actions, and the result lands in a live system instead of on a screen,” Afshar said.

That is why several sources argued that testing has to extend beyond the base model. Avi Pilcer, founder and CEO of Ultra Deep Tech, told AI Safety Watch in an email interview that independent red teams should have access to the model’s real tools rather than only a chat interface. Angelo Kevin Brown, a professor of criminology at Arkansas State University, also argued for third-party evaluation to reduce the risk of in-house blind spots.

Sachin Jade, chief product officer at Cyware, told AI Safety Watch in an email interview that release decisions should include long-horizon agent testing, least-privilege controls, staged rollout and human approval before sensitive actions. His response echoed a broader point from the other sources: the safety question is increasingly about the system around the model, not just the model itself.

Testing agents

Across the interviews, one recommendation came up repeatedly: test frontier AI as it will actually be deployed, not as an isolated model in a laboratory.

Fabrizio Di Carlo, managing director of ContrailRisks, told AI Safety Watch in an email interview that independent evaluators should examine a model together with the tools, permissions and safeguards it will actually have in use.

Di Carlo said those tests should include realistic cyberattack tasks, prompt injection, unauthorized actions, data leakage and attempts to defeat monitoring and shutdown mechanisms. Passing a laboratory evaluation, he said, cannot guarantee safe behavior once a system is connected to real tools and data.

Jaggi similarly called for independent red teams with no financial relationship to the developer, disclosure of evaluation results to regulators and containment testing aimed specifically at agentic deployments.

“Most current evaluations test the model,” Jaggi said. “The deployments that cause harm are agents.”

The distinction also complicates the idea that a model can pass a safety evaluation once and then be regarded as safe indefinitely. Models are updated. Tool access changes. Organizations connect them to different systems. Attackers also learn how to exploit weaknesses that were not apparent before launch.

Sharma said evaluations therefore should continue after deployment, with audits designed to detect behavioral changes as well as adversarial testing before a release.

“A model card describes what the model does before release,” Sharma said. “An audit at six months describes what it is doing differently.”

A real slowdown?

The question now is whether this week represents a genuine turning point or simply a temporary interruption in a race still driven by intense commercial competition.

There is evidence for both interpretations.

OpenAI’s decision to hold back GPT-6.1 Astra shows that internal safety thresholds can impose real costs on a product schedule. Other frontier developers have also become more explicit about staged releases, restricted access and additional testing when a model shows capabilities that raise new risks.

At the same time, the largest AI companies have not stopped training or releasing advanced systems. OpenAI launched other products this week even while Astra remained shelved. Competitive pressure to improve performance, cut costs and capture enterprise customers remains intense.

That makes the emerging pattern less a pause in artificial intelligence than a new constraint on how the most powerful forms of it reach the public.

The old model-development race looked relatively simple: train a more capable system, evaluate it and release it. The next phase may be less predictable. A model can clear capability benchmarks and still fail behavioral or security evaluations, forcing developers to redesign safeguards, limit access or abandon a release.

For AI companies, that could mean accepting that some technically impressive systems never become products in their original form.

The harder question is who gets to decide when the risk has fallen enough to try again.

Ramos said independent evaluators ultimately need real access and genuine authority to delay a deployment rather than merely advising the company that built it.

“The time to design brakes is before you’re driving 200 miles per hour,” Ramos said. “The goal isn’t to slow AI development. It’s to remain capable of governing it as it scales.”

Keep reading AI Safety Watch

Reporting on AI risk, security and governance. About the publication · Subscribe