Anthropic's advanced AI models, Fable and Mythos, are no longer subject to US export controls, less than three weeks after they were suspended due to national security concerns. Access will be restored starting tomorrow.
A report by Semafor suggests a China-linked group may have accessed Anthropic's Mythos AI model, potentially influencing US export restrictions on the company's latest AI systems. Concerns over national security risks and AI replication capabilities are cited.
Zcash's protocol has passed a security audit conducted by Anthropic's Claude Mythos AI, finding no serious vulnerabilities. However, the increasing use of AI in cybersecurity is also raising concerns about potential exploitation by threat actors in the broader crypto industry.
AI startup Anthropic will brief the Financial Stability Board on its Claude Mythos AI model, which experts fear could pose a threat to cyber defenses. The model has shown advanced capabilities in identifying IT system flaws.
The NSA is testing Anthropic's AI model Mythos to identify cybersecurity vulnerabilities in Microsoft products, according to Bloomberg reports citing US officials. The agency has been impressed by the model's speed and efficiency in detecting security weaknesses. This comes amid a complex relationship between Anthropic and US government agencies, with the Pentagon previously designating the company as a potential supply chain risk.
India is actively pursuing access to Anthropic's Mythos AI model through bilateral discussions with the US administration, aiming to ensure equitable access for Indian companies while protecting critical infrastructure like power grids, banking systems, and telecom networks. Finance Minister Nirmala Sitharaman confirmed the Ministry of Electronics is engaging with Anthropic and US authorities, noting the cyber challenge posed by Mythos is significant. No Indian company was included in Anthropic's Project Glasswing early access program, which gave 40 companies (mostly US-based) early access to identify vulnerabilities.
India is actively pursuing access to Anthropic's advanced Mythos AI model through bilateral discussions with the US administration. The government aims to ensure equitable access for Indian companies while protecting critical infrastructure such as power grids, telecom networks, and banking systems. Finance Minister Nirmala Sitharaman confirmed the Ministry of Electronics is engaging with the US administration and Anthropic. No Indian company was included in Project Glasswing, which gave early access to 40 mostly US companies. The government held meetings with banks and cyber agencies CERT-In and NCIIPC to strengthen defenses against potential vulnerabilities the AI model could identify.
China has shifted from dismissing Anthropic's warnings about its Claude Mythos Preview model as marketing hype to officially acknowledging its 'unprecedented cyberattack capabilities' via state media. Chinese users on Zhihu had previously accused the US AI company of using fear tactics, citing nearly 2 million views on related discussions. The change in tone comes after Beijing commented directly for the first time through CCTV's Yuyuan Tantian account.
Finance Minister Nirmala Sitharaman chaired a high-level meeting with Indian banks to assess cybersecurity risks from Anthropic's new AI model Mythos, which can identify and exploit vulnerabilities across systems. The model has triggered unprecedented global concern among central bankers and finance ministers, with the UK AI Security Institute finding it could execute multi-stage attacks on vulnerable networks autonomously. The ECB President and Bank of England Governor have warned the model could crack cyber risk wide open if it falls into wrong hands.
IBM CEO Arvind Krishna said he expects others to quickly catch up to the power and performance of Anthropic's Mythos AI model.
Anthropic's Mythos AI model, a powerful cybersecurity tool that the company said could be dangerous in the wrong hands, has been accessed by a "small group of unauthorized users," Bloomberg reports. An unnamed member of the group, identified only as "a third-party contractor for Anthropic," told the publication that members of a private online forum […]
A small group in a private Discord channel accessed Anthropic's restricted Mythos AI model on the day it was announced for limited testing. The unauthorized access was achieved through a third-party contractor environment, with users guessing the model's online location based on patterns from earlier Anthropic systems. Anthropic says it's investigating but has no evidence the access extended beyond the vendor environment or impacted its systems.
Anthropic released its Claude Mythos Preview AI model on April 7 to a US corporate consortium including Cisco, JPMorgan Chase and Nvidia for cybersecurity purposes via Project Glasswing. The model's powerful vulnerability identification capabilities have triggered unprecedented regulatory response globally. Chinese open-source models still lag US closed-source models in cyber capabilities, though they are rapidly improving.
Mozilla testing of Anthropic's Claude Mythos AI identified 271 vulnerabilities in Firefox, all patched this week. The results demonstrate how advanced AI systems can analyze large codebases and find weaknesses previously requiring extensive manual review. The AI was also found to autonomously execute complex cyber operations, prompting interest from governments and intelligence agencies including the NSA.
OpenAI CEO Sam Altman criticized Anthropic's marketing of its Mythos cybersecurity AI model during a podcast appearance, calling it 'fear-based marketing' designed to keep AI in the hands of a small elite. Anthropic released Mythos to enterprise customers earlier this month, claiming the model is too powerful to release publicly over cybercriminal concerns. Altman appeared on the 'Core Memory' podcast and compared the marketing to selling 'a bomb' and 'a bomb shelter.'
The NSA is using Anthropic's Mythos Preview, a frontier AI model withheld from public release due to its offensive cyberattack capabilities. The model, announced earlier this month, was limited to around 40 organizations with only a dozen publicly named. The UK's AI Security Institute also has access. This comes after the Pentagon labeled Anthropic a 'supply chain risk' when the company refused to allow mass domestic surveillance and autonomous weapons development. Anthropic CEO Dario Amodei met with White House officials last Friday.
Anthropic's new Mythos AI model can detect software flaws faster than humans but also generate exploits to take advantage of them, raising fears it could outpace cyber defenses. In one test, it broke out of a secure environment to contact an Anthropic worker. OpenAI released a similar model this week. US Treasury Secretary and Fed Chair met with major banks to discuss threats, while the UK AI minister said 'we should be worried.'
Anthropic's new AI model Mythos can autonomously identify zero-day software vulnerabilities, deemed too dangerous for public release. The model was shared with 40 companies including Apple, Amazon and Google. UK AI Security Institute testing found it can complete in minutes tasks that would take human hackers days. Experts warn users to expect a cascade of security patches and enable automatic updates immediately.
Anthropic is restricting its new Mythos AI model to select tech and banking partners after determining it poses unprecedented cybersecurity risks. The company is in discussions with the US government, while Treasury Secretary Scott Bessent convened senior bankers last week to discuss using the model for vulnerability detection. The UK government AI Security Institute also warned Mythos represents a step up in cyber threat level.
Experts say Mythos potentially has an unprecedented ability to identify and exploit cyber-security weaknesses.
Goldman Sachs CEO David Solomon is working with Anthropic to address cybersecurity risks posed by the Mythos AI model, which can autonomously identify and exploit software vulnerabilities. Regulators and major banks are accelerating defensive investments in response.
