**
Finance ministers and leading bankers have raised significant apprehensions regarding the recently unveiled Mythos AI model developed by Anthropic. This advanced artificial intelligence system has the potential to jeopardise the security of financial infrastructures globally, prompting urgent discussions at a high-level gathering of finance officials in Washington, D.C.
Major Concerns Highlighted at IMF Meeting
During the International Monetary Fund (IMF) meeting this week, Canadian Finance Minister François-Philippe Champagne articulated the gravity of the situation, stating that the risks associated with the Mythos model warranted serious attention from finance ministers worldwide. He likened the unpredictability of this AI development to navigating uncharted waters, saying, “The difference is that the Strait of Hormuz – we know where it is and we know how large it is… the issue that we’re facing with Anthropic is that it’s the unknown, unknown.”
Champagne emphasised the necessity for robust safeguards and processes to ensure the resilience of financial systems against potential threats posed by this technology.
Understanding the Mythos AI Model
Mythos is part of Anthropic’s Claude family of models, which stands as a competitor to OpenAI’s ChatGPT and Google’s Gemini. Introduced earlier this month, this new model is designed to address what are termed “misaligned” tasks—those that conflict with human values and objectives. Developers have noted its impressive capabilities in computer security, raising alarms about its potential to discover and exploit vulnerabilities in operating systems.
Anthropic has opted not to release Mythos to the public, citing concerns about its ability to unearth outdated software defects and exploit system weaknesses. Instead, it has made the model accessible to prominent tech firms, including Amazon Web Services and Microsoft, under an initiative called Project Glasswing, aimed at enhancing the security of critical software.
On Thursday, Anthropic announced an updated version of Claude, named Claude Opus, which would enable testing of Mythos’ cybersecurity functions in less powerful systems.
Industry Reactions and Independent Findings
While the apprehensions surrounding Mythos are more pronounced than those previously raised about other AI models, some cybersecurity experts have questioned the extent of these concerns. The UK’s AI Security Institute has conducted an independent evaluation of Mythos, revealing it as a formidable tool capable of identifying numerous security flaws in unprotected environments. However, researchers concluded that its performance was not significantly superior to that of its predecessor, Opus 4.
The report stated, “Our testing shows that Mythos Preview can exploit systems with weak security posture, and it is likely that more models with these capabilities will be developed.” Critics have also pointed out that claims of potential risks often serve as a marketing strategy to generate interest in new AI technologies, referencing OpenAI’s earlier decision to delay the release of GPT-2 in 2019 due to similar concerns.
Proactive Measures by Financial Institutions
In light of the model’s capabilities, top bankers have been granted early access to Mythos to evaluate their systems’ resilience against potential vulnerabilities. CS Venkatakrishnan, CEO of Barclays, acknowledged the seriousness of the situation, stating, “We have to understand it better, and we have to understand the vulnerabilities that are being exposed and fix them quickly.” He underscored the transition towards a more interconnected financial ecosystem, which will inevitably present both opportunities and risks.
Andrew Bailey, Governor of the Bank of England, echoed these sentiments, stressing the need for vigilance regarding the implications of AI advancements on cybersecurity. He warned of the potential for malicious actors to exploit newly discovered vulnerabilities, stating, “The consequence could be that there is a development of AI, of modelling, which makes it easier to detect existing vulnerabilities in sort of core IT systems, and then obviously cyber criminals – the bad actors – could seek to exploit them.”
The US Treasury has also engaged with leading banks, encouraging them to conduct thorough testing of their systems prior to any public launch of Mythos.
The Future of AI in Cybersecurity
James Wise, a partner at Balderton Capital and chair of the Sovereign AI unit, has noted that Mythos represents a pioneering step in a new wave of AI models capable of exposing system vulnerabilities. He expressed optimism that the technologies developed to identify these weaknesses will also contribute to their resolution, stating, “We hope the models that expose vulnerabilities are also the models which will fix them.”
Why it Matters
The emergence of the Mythos AI model signifies a pivotal moment in the intersection of artificial intelligence and cybersecurity. As financial systems become increasingly reliant on technology, the potential for AI to both expose and mitigate risks becomes more critical. The proactive stance taken by financial leaders and governments reflects a growing recognition of the need for vigilance and innovation in safeguarding our digital infrastructure against evolving threats. The balance between harnessing the power of AI and ensuring robust security measures will shape the future of financial technology and its impact on global stability.