German banks assess risks from Anthropic AI model amid cyber concerns

German banks and national authorities are assessing potential risks linked to a new artificial intelligence model developed by Anthropic, an official said on Thursday.

The concerns centre on whether the technology could increase the threat of cyberattacks targeting financial institutions.

Kolja Gabriel, a member of the executive board at the German Banking Association, said the group is actively consulting cybersecurity experts across its member banks.

He added that discussions are also taking place with Germany’s finance ministry and other authorities, according to Reuters.

Coordination with regulators and authorities

The ongoing talks include participation from the Bundesbank and Germany’s financial regulator BaFin.

BaFin confirmed that it is maintaining regular exchanges with national, European and international stakeholders.

The regulator stressed the need for preparedness within the financial sector.

Mythos raises concerns among cybersecurity experts

Anthropic’s model, known as Mythos, has drawn attention from cybersecurity experts who believe it poses significant challenges for banks, particularly those relying on legacy technology systems.

These concerns have also prompted regulatory interest in Britain and the United States.

“Mythos is being used in a controlled manner by IT security firms to close potential vulnerabilities as quickly as possible. We expect a series of software updates shortly and are closely monitoring developments,” Gabriel said in an emailed statement to Reuters.

He is responsible for technology and innovation at the association.

Supervisors at the European Central Bank are preparing to question bankers about risks linked to the model.

Limited rollout and industry testing

Anthropic has stated that its current version, Claude Mythos Preview, will not be released for general use.

Instead, the company has launched an initiative called Project Glasswing.

The programme invites major technology firms, cybersecurity vendors and organisations including JPMorgan Chase to privately evaluate the model and strengthen defences.

This controlled rollout reflects broader caution within the industry as stakeholders seek to understand the implications of advanced AI systems.

Anthropic’s rising influence in AI landscape

The scrutiny comes at a time when Anthropic is gaining prominence in the global artificial intelligence sector.

The company has emerged as a major player following a significant funding round that valued it at $380 billion.

According to Dave Kasten of Palisade Research, Anthropic has built a strong position by focusing on security and reliability.

He noted that the company’s approach to “constitutional integrity” has created a structural advantage over competitors.

Kasten highlighted the model’s ability to detect vulnerabilities through advanced reasoning, rather than relying solely on historical patterns.

He cited an example where the system identified a previously undetected flaw in a FreeBSD implementation, demonstrating its potential for proactive cybersecurity.

Strategic importance and policy tensions

Despite concerns, interest in Mythos extends beyond Europe.

A Politico report indicated that US agencies and congressional officials continue to engage with the model’s capabilities, even after restrictions were placed on working with Anthropic.

The report said the Commerce Department’s Center for AI Standards and Innovation is testing Mythos’ cybersecurity features.

Staff across multiple congressional committees have also sought briefings on the model.

The situation highlights a broader tension.

While the Pentagon has labelled Anthropic a supply-chain risk and cut business ties, parts of the US government appear reluctant to ignore a technology seen as strategically important.

The extent of this engagement remains unclear.

However, the continued interest suggests that advanced AI tools like Mythos are becoming increasingly central to cybersecurity and national security considerations, even amid regulatory caution.

The post German banks assess risks from Anthropic AI model amid cyber concerns appeared first on Invezz