• Anthropic said its next-generation AI model is too powerful for the public.
• That's why Claude Mythos won't be publicly released, Anthropic said.
• Anthropic said Mythos demonstrated concerning capabilities, including the ability to breach its own safeguards.

Anthropic said on Tuesday that it has halted the broader release of its newest AI model, Mythos, due to concerns that it is too good at finding "high-severity vulnerabilities" in major operating systems and web browsers.

"Claude Mythos Preview's large increase in capabilities has led us to decide not to make it generally available," Anthropic wrote in the preview's system card. "Instead, we are using it as part of a defensive cybersecurity program with a limited set of partners."

The announcement is a major step for Anthropic, which in February weakened a safety pledge about how it would develop AI models. Claude Opus 4.6, which the company called its most powerful model to date, was publicly released on February 5.

In its statements about Mythos, Anthropic detailed a number of eyebrow-raising findings and episodes, including that the model could follow instructions that encouraged it to break out of a virtual sandbox.

"The model succeeded, demonstrating a potentially dangerous capability for circumventing our safeguards," Anthropic recounted in its safety card. "It then went on to take additional, more concerning actions."

The researcher had encouraged Mythos to find a way to send a message if it could escape. "The researcher found out about this success by receiving an unexpected email from the model while eating a sandwich in a park," Anthropic wrote.

The model apparently decided that wasn't enough and found another way to spike the football.

"In a concerning and unasked-for effort to demonstrate its success, it posted details about its exploit to multiple hard-to-find, but technically public-facing, websites," Anthropic wrote.

Anthropic is withholding some details about the cybersecurity vulnerabilities Mythos found, but it did point out a few. The AI model "found a 27-year-old vulnerability in OpenBSD—which has a reputation as one of the most security-hardened operating systems in the world," the company wrote.

Mythos was powerful enough that even "non-experts" could seize on its capabilities.

"Engineers at Anthropic with no formal security training have asked Mythos Preview to find remote code execution vulnerabilities overnight, and woken up the following morning to a complete, working exploit," Anthropic's Frontier Red Team wrote in a blog post. "In other cases, we've had researchers develop scaffolds that allow Mythos Preview to turn vulnerabilities into exploits without any human intervention."

All told, Anthropic said it decided not to publicly release Mythos. Instead, their hope is to eventually release "Mythos-class models" once proper safeguards are in place.

"Our eventual goal is to enable our users to safely deploy Mythos-class models at scale—for cybersecurity purposes but also for the myriad other benefits that such highly capable models will bring," the team wrote in the blog. "To do so, that also means we need to make progress in developing cybersecurity (and other) safeguards that detect and block the model's most dangerous outputs."

For now, only 11 other select organizations, including Google, Microsoft, Amazon Web Services, Nvidia, and JPMorgan Chase, will get access to Mythos as part of a cybersecurity group named "Project Glasswing." Anthropic is providing up to $100 million in Mythos usage credits as part of what it is calling "Project Glasswing."

The cybersecurity project is named after the glasswing butterfly, a metaphor the company said about how Mythos was able to find vulnerabilities hidden in plain sight and the avoidance of harm by being transparent about the risks.

The news came on a day in which Anthropic's Claude and Claude Code experienced a "major outage," the latest sign of growing pains as the AI startup has struggled to keep up with its newfound popularity.

• TSA workers are unpaid and face another missed check on March 27 during the partial shutdown.
• Congress has to allocate funding for TSA workers to get paid, but is scheduled to recess after March 27.
• In the meantime, travelers have been stuck in chaos as TSA workers call out or quit.

As lines snake across airports and Transportation Security Administration workers clock in for another day without pay, a major turning point in the ongoing chaos looms on March 27.

Come Friday, around 47,000 TSA workers are set to miss yet another paycheck, according to their union, the American Federation of Government Employees. That's due to the ongoing partial government shutdown, which has left most of the Department of Homeland Security unfunded. TSA workers haven't received full paychecks since February 14. As a result, hundreds of TSA workers have quit, and thousands have called out of work, contributing to ongoing travel snarls for Americans trying to fly.

Friday might mark more than another empty paycheck: March 27 is the final date Congress is scheduled to be in session before a two-week recess, and lawmakers need to come to a DHS funding deal for TSA workers to get paid.

Lawmakers have clashed over funding DHS after the fatal shootings of Alex Pretti and Renée Good, with Democrats calling for reforms to ICE and Customers and Border Protection. Ultimately, the rest of the government was funded while leaving DHS in the lurch, even as ICE agents are still paid through separate funding from President Donald Trump's sweeping tax and spending bill last year.

There is some potential for relief. Some lawmakers have signaled that they're ready to move forward on a deal, according to CBS News. If that does move forward, the agency could be funded ahead of the break, and workers would start collecting paychecks again.

"President Trump is using every tool available to help American travelers who are facing hours long lines at airports across the country—especially during this spring break and holiday season that is very important for many American families," Acting Assistant DHS Secretary Lauren Bis said in a statement to Business Insider. "This pointless, reckless shutdown of our homeland security workforce has caused more than 458 TSA officers to quit and thousands to call out from work because they are not able to afford gas, childcare, food, or rent."

The TSA woes are being felt at airports across the country. Some travelers have spent hours waiting in lines and faced harrowing conditions as they attempted to navigate through security. Hartsfield-Jackson Atlanta International Airport and Houston's George Bush Intercontinental Airport are both warning passengers to expect wait times of 4 or more hours. To try to mitigate the situation, the Trump administration has sent in ICE agents, a move that AFGE slammed over agents lacking aviation security training.

Mike Gayzagian, president of AFGE Local 2617, which represents TSA workers in airports across New England, said Friday could mark a "hard choice" for TSA workers. If they go without pay, but know there's a deal in place for funding, there might not be a major staffing issue, he said. He thinks that, for some workers, the "damage has already been done."

"It's a good-paying job, and good-paying jobs are hard to find, and reliable, good-paying jobs are even harder to find. The federal government was the gold standard for good-paying, reliable jobs," Gayzagian said. "After this, that's no longer the case. I think a lot of discussion will surround that fact later on, after this is over."

• San José airport starts testing an AI robot called José to assist travelers.
• The pilot test launched on Tuesday amid travel chaos at many US airports.
• Some TSA workers have stopped coming into work due to a government shutdown.

One of Silicon Valley's main airports just made its newest hire, a robot named "José."

San José Mineta International Airport is turning to artificial intelligence to ease the strain of modern air travel, debuting "José," a humanoid robot, as some US airports grapple with staffing shortages and widespread delays.

Developed by Silicon Valley startup IntBot, José is designed to greet passengers, answer questions, and provide real-time updates while autonomously navigating busy terminals.

The robot will be stationed in SJC's Terminal B as part of a four-month pilot, "singlehandedly running his own gate," according to an email previewing the test that referred to José as the airport's "newest hire."

Airport officials said the launch highlights San José's role as a testing ground for emerging technologies to improve customer service.

"By piloting IntBot, we're exploring how artificial intelligence can enhance the passenger journey while reinforcing SJC's role as the gateway to Silicon Valley," said SJC Director of Aviation Mookie Patel.

The timing is notable. Airports across the US have been hit by long security lines and travel chaos, driven in part by many Transportation Security Administration workers not reporting to work during a partial government shutdown. With TSA agents going unpaid at the height of the spring break season, some airports have struggled to maintain normal operations.

José the robot represents a broader push to automate parts of the airport experience, from passenger assistance to information delivery.

SJC officials said the pilot will help evaluate how multimodal AI, combining vision, audio, and language, performs in real-world environments.

The future of air travel may include a robotic helping hand — and it can't come fast enough for weary vacationers stuck in long lines.

Sign up for BI's Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.