In a recent earnings call, David Solomon, the CEO of Goldman Sachs, expressed acute awareness of the cybersecurity risks associated with Anthropic’s latest AI innovation, the Mythos model. Solomon’s remarks follow a high-profile meeting in Washington with top bankers, including U.S. Treasury Secretary Scott Bessent, where concerns about the evolving capabilities of artificial intelligence in cyber threats were at the forefront. As financial institutions grapple with the implications of advanced AI technologies, Solomon emphasised the bank’s proactive stance on enhancing its cybersecurity measures.
A New Era of Cyber Threats
The rapid evolution of artificial intelligence, particularly large language models (LLMs), has raised alarms within the banking sector. Solomon highlighted that these advancements necessitate vigilant monitoring, particularly as institutions like Goldman Sachs strive to bolster their defences against increasingly sophisticated cyberattacks. In a statement, he remarked, “We’re hyper-aware of the enhanced capabilities of these new models with the help of the US government and the model publishers,” affirming the bank’s commitment to staying ahead of potential threats.
Anthropic’s Mythos model, which was recently flagged for its unprecedented ability to identify and exploit software vulnerabilities, is reportedly a significant concern for financial stability. The company’s blog post warned that current AI models have reached a level of sophistication that enables them to outperform many human experts in pinpointing weaknesses within IT infrastructure. Such developments could have dire repercussions for economies, public safety, and national security, as articulated by Anthropic.
Collaborative Efforts to Mitigate Risks
In light of these warnings, Solomon confirmed that Goldman Sachs is actively collaborating with Anthropic and various security vendors to leverage the capabilities of Mythos. “We have the model. We’re working closely with Anthropic and all of our security vendors to kind of harness frontier capabilities wherever it’s possible,” he said. This collaboration is part of a broader strategy aimed at strengthening the bank’s cyber resilience, with Solomon noting that investments in this area are being accelerated.
The urgency of these measures was underscored by the recent gathering of major American bankers in Washington. This meeting aimed to assess the implications of Mythos and its potential to disrupt the operations of systemically important banks—financial institutions whose failure could trigger widespread economic instability.
The Global Response to Mythos
The implications of Mythos are not confined to the US. The UK government’s AI Security Institute (AISI) has also raised alarms, labelling Mythos a significant escalation in cyber threat capability. AISI’s analysis revealed that Mythos can execute multi-step cyber-attack simulations, tasks that would conventionally require human expertise and considerable time. In tests conducted by AISI, Mythos successfully completed a complex 32-step simulation in just three attempts, showcasing its potential for autonomous cyber operations.
While AISI highlighted Mythos’s ability to target less-defended IT systems, it noted the limitations of its testing framework, which may not fully capture the model’s capacity against more robust defensive measures. The institute concluded with a stark warning that future iterations of AI will likely enhance these capabilities, urging immediate investment in cyber defence mechanisms.
Regulatory Scrutiny on the Horizon
In the coming weeks, UK regulators are set to convene discussions focused on the risks posed by Mythos with leading bank executives and government officials. The Cross Market Operational Resilience Group (CMorg)—comprising chief executives and representatives from key regulatory bodies such as the Treasury and the Bank of England—will explore the broader implications of these AI developments.
As discussions unfold, the Bank of England is managing communications regarding these deliberations, though it has thus far refrained from commenting on the specifics of the situation.
Why it Matters
The emergence of Anthropic’s Mythos model epitomises the dual-edged nature of AI advancements in the financial sector. While these technologies promise enhanced operational efficiencies, they simultaneously introduce unprecedented risks that could jeopardise the integrity of financial systems worldwide. By acknowledging these threats and taking proactive measures, institutions like Goldman Sachs are not only safeguarding their operations but also setting a precedent for the industry. As the landscape evolves, the imperative for robust cyber resilience will become increasingly paramount, underscoring the need for a collaborative approach to securing our digital future.
