In a significant development for the financial sector, David Solomon, the chief executive of Goldman Sachs, has expressed heightened concern regarding the cybersecurity challenges posed by Anthropic’s advanced AI model, Mythos. In a recent earnings call, Solomon disclosed that the investment bank is actively collaborating with Anthropic to bolster its cyber defences amid warnings from the tech firm about the unprecedented risks associated with this new technology.
Increased Vigilance in the Face of AI Advancements
Goldman Sachs has been closely monitoring advancements in artificial intelligence, particularly large language models (LLMs), as part of its strategic initiative to enhance cybersecurity measures. During the earnings call, Solomon emphasised the rapid progression in AI capabilities and conveyed his acute awareness of these developments. “Obviously the LLMs are making rapid progress, and we’re hyper-aware of the enhanced capabilities of these new models with the help of the US government and the model publishers,” he articulated to analysts.
This comes in light of statements from Anthropic, creators of the Claude AI family, which recently highlighted the alarming potential of Mythos. According to Anthropic, the model’s ability to identify and exploit IT vulnerabilities could greatly surpass that of all but the most adept human hackers. In a blog post, the firm stated, “AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.” The implications of this capability, they warned, could be severe for economic stability and national security.
Collaborative Efforts to Strengthen Cyber Resilience
In response to these threats, Solomon confirmed that Goldman Sachs possesses the Mythos model and is working in tandem with Anthropic and various security vendors to leverage its advanced capabilities. “We’re aware of Mythos and its capabilities… We’re working closely with Anthropic and all of our security vendors to harness frontier capabilities wherever it’s possible,” he stated, underscoring the bank’s commitment to enhancing its cyber and infrastructure resilience.
The urgency of these discussions was further underscored by a recent meeting convened by US Treasury Secretary Scott Bessent, which included Solomon and other leaders from systemically important banks. The focus of the meeting centred on the potential fallout from disruptions at these institutions, which regulators believe could threaten overall financial stability.
UK Government Responds to Emerging Threats
As concerns regarding Mythos escalate, the UK government’s AI Security Institute (AISI) has also issued a warning, characterising the model as a significant advancement in cyber threat capabilities. AISI’s findings suggest that Mythos can execute complex cyber-attack simulations autonomously, identifying weaknesses in IT systems much faster than human professionals. They noted that Mythos successfully navigated a 32-step simulation of a cyber-attack, achieving success in three out of ten attempts.
The institute noted that while Mythos appears adept at targeting smaller, poorly defended systems, the true extent of its capabilities against well-secured networks remains unclear due to limitations in testing environments. Their analysis concluded with a stark reminder that future AI models are likely to improve upon Mythos, making immediate investments in cybersecurity crucial.
In response, UK regulators are planning to discuss the implications of Mythos with banking executives and government officials in the upcoming weeks. The Cross Market Operational Resilience Group (CMorg), which includes chief executives and representatives from the Treasury, Bank of England, Financial Conduct Authority, and the National Cyber Security Centre, is set to convene soon.
Why it Matters
The escalating sophistication of AI models such as Mythos presents a formidable challenge for financial institutions and regulators alike. As banks like Goldman Sachs strive to navigate this evolving landscape, the emphasis on robust cybersecurity measures has never been more critical. The collaboration between financial giants and tech firms could reshape the approach to cybersecurity, but it also highlights the urgent need for comprehensive strategies to mitigate risks posed by advanced artificial intelligence. The potential for these models to undermine financial stability underscores the necessity for proactive investment in cyber defence mechanisms, positioning the sector to better withstand future threats.
