Goldman Sachs 'Hyperaware' as it Tests Mythos for Defense

Goldman Sachs 'Hyperaware' as it Tests Mythos for Defense CEO Solomon Says Bank is Working with Anthropic, Vendors on Controls Rashmi Ramesh (rashmiramesh_) • April 14, 2026     Share Post Share Credit Eligible Get Permission Image: Shutterstock Goldman Sachs is putting to work Anthropic's most powerful and most restricted artificial intelligence model to understand what it can do to the financial system's defenses. See Also: AI Impersonation Is the New Arms Race-Is Your Workforce Ready? Speaking at the bank's Monday earnings call, CEO David Solomon said the financial institution is testing Mythos, a new AI model that has raised concerns about enhanced cyberattacks. "We're hyper-aware of the enhanced capabilities of these new models," he said. "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." The call came days after Department of the Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned Wall Street leaders to a meeting over concerns that Mythos would usher in an era of greater cyber risk. Solomon did not frame the gathering as extraordinary. "It is not the first meeting that that group has gone over to Treasury to talk about cybersecurity risks over a number of years," he said. "This is something the industry is focused on. It's something we're focused on. And there's nothing new in that focus." Anthropic launched Mythos earlier this month but restricted access to a few dozen hand-picked companies that form what it dubs Project Glasswing, a group meant to use Mythos to root out security flaws before hackers get ahold of similar functionality. Anthropic researchers said Mythos identified thousands of previously unknown vulnerabilities across every major operating system and every major web browser over just a few weeks of testing (see: Anthropic Calls Its New Model Too Dangerous to Release). A concern driving attention from Wall Street and Washington is not just what the model can find, but who could use it. Analysts say Mythos represents a significant transition with its ability to autonomously generate working exploits without human guidance and enable autonomous attacks at speed and scale. Great Britain's AI Security Institute, which evaluated Mythos independently, found it could execute multi-stage attacks on vulnerable networks and discover and exploit flaws without human direction - tasks that would take human professionals days. Mythos was the first model to complete a 32-step simulated corporate network attack that the institute built to test multi-step intrusion scenarios, succeeding in three out of ten attempts. The institute said it could not determine whether Mythos could breach well-defended systems since its test environments lacked active defenses such as endpoint detection and real-time incident response. Goldman Sachs has been building AI-powered automation with Anthropic for more than a year. Solomon told investors in the call that "with the help of the U.S. government and the model publishers, we are very focused on supplementing our cyber and infrastructure resilience. And this is part of our ongoing capabilities that we have been investing in and are accelerating our investment in." Goldman Sachs CIO Marco Argenti in February said that embedded Anthropic engineers had spent six months at the firm co-developing autonomous agents to handle trade accounting and client onboarding, CNBC reported. He added that the bank is likely to automate other parts of its operations. "Whenever you have acceleration of your technology, there are going to be bumps, and there are going to be risk issues," Solomon said in response to a separate question on the earnings call. "But the power of the technology, the ability to use it in an enterprise, to remake processes, to create efficiency, and also create more capacity to invest in growth - I can't find a CEO that's not talking about that."