Dutch MEPs angered by Europe’s slow approach to AI, hacking concerns
EU lawmakers are warning that Europe is being left out of testing a powerful new artificial intelligence system capable of rapidly identifying cybersecurity vulnerabilities, raising fears that governments and companies in the bloc are unprepared for large-scale hacking threats, AD reported.
The AI model, called “Mythos,” was developed by the U.S. company Anthropic and is currently being tested by more than 40 American companies, including Google, Apple, and Microsoft, along with some British banks. EU institutions and companies are not included in the program.
Anthropic has reportedly not released the system publicly because of concerns about its capabilities. Instead, selected organizations are using it to identify weaknesses in their own systems ahead of possible cyberattacks. European Parliament members said the exclusion prevents the EU from assessing its vulnerabilities.
VVD Member of the European Parliament Bart Groothuis, who co-signed a joint letter to the European Commission, said participants in the testing program were alarmed by what the technology could do.
“I spoke with several parties that participated in the program. They are completely shocked by what this machine is capable of. Europe is watching and taking a wait-and-see approach. I think that is the dumbest thing you can do,” Groothuis told AD.
GroenLinks-PvdA (soon to be PRO) MEP Kim van Sparrentak said Europe needed direct access to systems like Mythos to prepare defenses. “We want to be able to check what weaknesses we have ourselves. Every system has its weaknesses, and we must at least be able to take precautionary measures. In a few months, there will already be many systems like this, according to most experts,” she said.
She warned the technology could be used to attack governments and institutions. “The AI can hack companies, institutions, ministries, and so on,” Van Sparrentak said.
Lawmakers also raised concerns about similar Chinese open-source AI systems that are publicly available and difficult to regulate. Groothuis said some were already nearing Mythos’ capabilities.
“This kind of creator is already close to Mythos, and there is no company behind it. It is public and cannot be regulated,” he said.
Groothuis said one of the biggest risks was that autonomous AI agents could independently use the technology for cyberattacks.
“The extra frightening thing about the latest AI software from Anthropic is that AI agents can use it on their own. The AI can hack companies, institutions, ministries, and so on. Criminal states and bad actors then have a golden opportunity in their hands. That is very troubling for everyone,” he told AD.
He warned that the consequences could spread far beyond governments and corporations. “Think of Odido, but on a massive scale. It could happen on such a large scale that you think: wow. All the software we have becomes vulnerable. That is why we must prepare ourselves. We need to act quickly. I miss that urgency in politics,” Groothuis said.
He also criticized what he described as a lack of urgency within EU institutions. “While the White House is personally involved in this, there is a kind of radio silence in the European Commission,” he said.
“That makes me very nervous. The newspapers are full of it, but you do not hear politics about it. Instead, they focus on topics like the Oostvaardersplassen or the wolf, but this is something very serious.” Van Sparrentak said Europe should not become dependent on U.S.-controlled AI technology.
“You do not want this to be used as a weapon by Americans,” she said. During a parliamentary hearing Wednesday, the European Commission argued that existing EU regulations, if properly implemented, could protect organizations against attacks involving systems like Mythos.
Lawmakers disputed that assessment. “The whole point is that the people developing these systems aren't concerned about laws or institutions. They see themselves as almighty. The longer we spend talking about laws, the more time we lose for real solutions,” said Dirk Gotink of the NSC.