The rollout of a powerful new AI model, Mythos, an advanced system developed by Anthropic, has hit a roadblock after the White House opposed plans to expand its access, according to The Wall Street Journal. The company wanted to open the model to more organisations, but the US government isn’t on board, at least not yet.
White House pushes back on Anthropic’s Mythos plan
Anthropic recently proposed giving access to around 70 more companies and organisations. That would have taken the total number of users to about 120. But officials in the White House seemed against that move.
According to WSJ, citing people familiar with the matter, the administration is worried about security risks. Mythos is not just another AI tool, it is capable of finding weaknesses in software systems, something that could be misused for cyberattacks or large-scale disruption online.
There were also concerns about resources. Some officials questioned whether Anthropic has enough computing power to handle that many users without affecting the government’s own access to the system.
Why Mythos is raising alarms
The White House is closely involved in how Mythos is being rolled out because of the risks it carries. In recent weeks, the model’s ability to detect and exploit software bugs has already made both government agencies and private companies uneasy.
Cybersecurity experts, according to WSJ, say tools like Mythos, along with similar systems being developed by companies like OpenAI and Google, are becoming so powerful that they could make it easier for bad actors to launch attacks.
At the same time, these tools are also useful for defence. They can help researchers find and fix bugs before hackers exploit them. But the sheer number of vulnerabilities being uncovered is becoming a challenge in itself.
This latest disagreement shows that the relationship between Anthropic and the White House is still not smooth. Both sides have tried to cool things down in recent months, but tensions are clearly visible.
Earlier, the administration had even considered cutting ties with the company over a dispute with the Pentagon about how its AI tools could be used by the military. That issue is now entangled in two ongoing court cases.
There are also political undertones. The administration has criticised Anthropic for its links to organisations that support stricter AI regulation and for employing several former officials from the previous government.
Limited rollout for now
So far, Anthropic has given access to about 50 companies and organisations, mainly those handling critical infrastructure. Some US government agencies are already using the model, and the administration is looking to expand its own access.
There are no plans yet to release Mythos to the general public. A White House official said the government is trying to strike a balance, allowing innovation to move forward, while making sure such powerful tools are introduced safely.
On the other hand, adding to the worries, Anthropic recently said it is looking into possible unauthorised access to Mythos. That has raised fresh concerns about what could happen if such a system falls into the wrong hands, especially given how quickly it can identify software flaws.
Computing power questions
Some White House advisers, including venture capitalist David Sacks, have suggested that Anthropic may not yet have the infrastructure needed for a wider rollout. The company has signed deals with firms like Amazon, Google and Broadcom to boost its capacity. But those projects will take time to fully come online.
In the meantime, AI companies often prioritise who gets access to computing power, and governments are usually at the front of that line. For now, Mythos remains tightly controlled.