Canada regulator warns major banks on Anthropic AI cyber risks – report

Canada's national banking watchdog alerted the country's fiscal institutions connected risks tied to Anthropic's Claude Mythos and different precocious AI models, reported Reuters.

According to the regulator, the exertion could summation cyber threats and trim the clip disposable to observe and spot vulnerabilities.

The Office of the Superintendent of Financial Institutions (OSFI) sent the connection to main exertion officers, main accusation information officers and main hazard officers crossed the fiscal sector, including ample banks and insurers, according to documents Reuters obtained done an access-to-information request.

Authorities successful respective jurisdictions are examining cybersecurity concerns linked to Anthropic's frontier AI exemplary Mythos.

The exemplary has been described arsenic highly susceptible of identifying and exploiting cybersecurity vulnerabilities, creating challenges for banks and their older exertion systems.

"Advanced artificial quality models, specified arsenic Anthropic Claude Mythos, ⁠significantly compress the timeframe for effectual hazard mitigation," OSFI said successful an email.

"Accordingly, this bulletin is grounded successful our existing guidance and outlines dependable practices that institutions tin follow to heighten the velocity and effectiveness of hazard identification, mitigation and response."

Recognition of the risks associated with Mythos by OSFI whitethorn pb Canadian banks, insurers and different regulated institutions to enactment much resources into exertion aimed astatine protecting clients from cyber threats, noted the quality agency.

In a connection to Reuters, the regulator said: "OSFI takes a technology‑neutral, risk‑focused attack to emerging technologies, including precocious artificial quality models specified arsenic ​Mythos. Our absorption is not the exertion itself, but however federally regulated fiscal institutions govern and negociate the risks associated with its use."

In aboriginal April, Canadian banking executives met regulators to sermon risks linked to Mythos, soon aft US Treasury Secretary Scott Bessent and then-Federal Reserve Chair ‌Jerome Powell ⁠held an urgent gathering with slope main executives to pass astir cyber risks connected to Anthropic's latest AI model.

OSFI sent the email to institution executives successful April.

The cyber capabilities of immoderate frontier AI systems are considered that entree has been restricted, with presently excluded from Mythos.

Three of Canada's large six banks, Royal Bank of Canada, TD Bank and BMO, person acceptable retired plans to marque millions from AI investments arsenic they moved from proceedings projects to uses specified arsenic chatbots, interior tools and reducing dependence connected third-party tools.