Banks and regulators around the world are sounding urgent warnings after Anthropic, the AI company behind the Claude chatbot, disclosed that its most advanced model, Mythos, can identify thousands of severe security flaws across major operating systems and web browsers. Internal testing reportedly found many long‑undetected weaknesses, including “zero‑day” vulnerabilities that require immediate fixes, prompting Anthropic to restrict public access and share the model only with a small defensive coalition and select partners.
Anthropic has made Mythos available to roughly a dozen partners — including Microsoft, Amazon Web Services, Apple, Cisco and the Linux Foundation — and to more than 40 other organisations, among them some US banks, according to the company. It has also committed US$100 million in usage credits and US$4 million in open‑source grants to help find and patch bugs. Reuters has reported that, as of mid‑April 2026, Anthropic had not granted Mythos access to banks in Australia, the United Kingdom or Europe. Anthropic is also investigating a Bloomberg report that a small group of unauthorised users may have accessed Mythos; the company says it is probing those claims and has not alleged malicious intent.
The concern is straightforward: banks hold money and run critical systems that often include decades‑old legacy software. Regulators at the International Monetary Fund spring meeting in Washington in April 2026 joined policymakers in warning that AI tools with advanced cyber capabilities could widen the attack surface for financially motivated criminals. For individuals, however, the immediate risk remains limited by consumer protections in many countries. In Australia, for example, the first A$250,000 of a depositor’s funds is covered by the government’s Financial Claims Scheme, and the Australian Securities and Investments Commission requires banks to investigate and reimburse customers for fraud when they are not at fault.
Experts advise practical precautions: keep computers and smartphones updated with the latest operating systems and banking apps, because vendors will be issuing patches as new vulnerabilities are found. Be especially wary of phishing attempts by email and SMS that try to steal banking credentials. With defence more difficult than attack, the software supply chain and long‑running systems present an ongoing race to find and fix flaws before they are exploited. Recent examples underline that point: the European Union’s age‑verification app, released on April 15, 2026, had security holes discovered within hours. Ambitious projects such as the Beneficial AI Foundation’s effort to formally verify Signal aim to show how mathematical proofs and better tools might reduce bugs, but for now defenders remain on the back foot.
Toby Walsh, University of New South Wales, Sydney. This article is republished from The Conversation under a Creative Commons licence.
Original Source: https://nenow.in/science-technology/why-the-worlds-banks-are-so-worried-about-anthropics-latest-ai-model.html
Category: Opinion,Tech,Top News
Tags:
Publish Date: 2026-04-26 23:25:00
