BRUSSELS — U.S. artificial intelligence leader Anthropic is facing growing criticism over its decision to keep a new model with supreme hacking abilities closed off from European regulators.

The company’s actions will face scrutiny in the European Parliament on Wednesday, where officials from the European Commission and the bloc’s cyber agency ENISA will discuss the dangers of the Mythos model. Anthropic won’t join the hearing, saying it was “unable to accept” the invitation “at short notice.”

The fact that Anthropic said it will skip the meeting is “extremely worrying,” said Dutch European Parliament lawmaker Kim van Sparrentak, who called the debate.

Wednesday’s hearing comes as Brussels and EU capitals are losing patience with the fact that Anthropic has not shared Mythos beyond a small circle of mostly American organizations.

Anthropic said early April its new model outperformed most humans in finding and exploiting software glitches and opted to shield the technology from wider use over worries that it could fall into the wrong hands. But nearly a month after the U.S. company announced Mythos to the world, European regulators are still struggling to test the technology and prepare for its impacts.

The European Commission dialed up its rhetoric on Tuesday, with Commission spokesperson Thomas Regnier telling POLITICO: “Once the enforcement powers of the AI Office start in August 2026, we will ensure to receive, if needed, model access.”

And on Monday, 30 members of the European Parliament called on Commission tech chief Henna Virkkunen to come up with a “European mitigation plan” to protect Europe’s “crown jewels” against the hacking risks, including by pushing for the bloc’s cyber agency ENISA to gain access to Mythos.

“Europe is not at the table,” said Bart Groothuis, a Dutch liberal lawmaker and one of the letter’s signatories. “I think that we should shake the tree much, much harder.”

Spain has asked the European Commission’s AI Office for information and more coordination on Mythos, Alberto Gago, director general of the Spanish AI authority AESIA, revealed on stage at POLITICO’s AI & Tech Week event in Brussels on Tuesday.

And the topic could pop up at the next meeting of the AI Board, the gathering of AI authorities of all 27 member countries, Gago said.

“It’s absolutely vital that AI companies release models like Mythos safely,” van Sparrentak said. “We cannot rely solely on the goodwill of companies for this.”

Banks on edge

Europe’s banking sector is also scrambling to assess and mitigate the risks to financial markets and systems if AI systems with advanced hacking abilities run rampant. While JPMorganChase is on the list of 12 named organizations with which Anthropic has shared Mythos, no European banks are.

The situation in Brussels is a sharp contrast to the U.S., where it was announced Tuesday that top AIfirms Microsoft, xAI and Google DeepMind had joined OpenAI and Anthropic in inking deals that will allow the government to vet their AI models for national security risks ahead of release.

Both the EU’s Economy Commissioner Valdis Dombrovskis and Eurogroup President Kyriakos Pierrakakis on Monday called for clarity to help the continent’s banks protect themselves.

The Commission is in discussions for “European companies and notably the banking sector” to be able to conduct cyber resilience testing using Mythos, Dombrovskis said after a meeting of the Eurogroup, the finance ministers of the countries of the eurozone.

For Pierrakakis, who is also Greece’s finance minister, the geopolitical situation is making the task more difficult: “The challenge here is that technologies like AI necessitate international governance frameworks at a moment where multilateralism is challenged,” he said in the same press conference.

But the scare around Mythos comes amid rocky EU-U.S. relations, with clashes over a range of issues from military support to trade flows and online freedom of speech.

U.S. Senator Gary Peters, a Democrat from Michigan and ranking member of the U.S. Senate’s homeland security committee, told POLITICO’s AI & Tech Week event on Tuesday he wants to “explore” why Anthropic’s superhacking model Mythos has not been shared with EU regulators for review.

“I’d love to explore why it’s not being provided to folks here in Europe as well. Because you’ve got to be part of the solution. There’s just no question about it in my mind,” Peters said.

Brussels and Washington plan to set up a new forum to discuss technology regulation, they said at the start of April, but details remain sparse and there is a deep rift between the two sides’ positions on tech rules from data regulation to content moderation.

According to Pierrakakis, Europe doesn’t have “the luxury of not trying to establish channels of communication with the United States, both on this topic [of superhacking AI] and on other topics as well, because frankly, the changes that we will see coming from AI are going to be significant.”