Billion-dollar bets, shock losses & new AI rules in finance.

In a major signal of AI’s growing strategic importance, payments titan Stripe has agreed to acquire OpenRouter, an AI model-routing platform, for over $7 billion ([1]). OpenRouter’s service can direct requests across more than 400 large language models and has amassed millions of developer users ([2]). By buying this "Stripe for AI" gateway, Stripe aims to embed itself at the core of the AI ecosystem, effectively becoming the billing and integration rail for a vast array of AI services – much as it did for online payments. The deal’s 5x premium over OpenRouter’s valuation just three months ago highlights both the fervor and the stakes in the race to own key AI infrastructure.

Incumbent financial powerhouses are also accelerating their AI investments. Goldman Sachs recently launched an AI-driven investment platform (dubbed AlphaAI) to identify opportunities from the boom in AI-powered businesses. The firm even created a new leadership role of "Chairman of AI" in its asset management arm, reflecting a top-down commitment to embedding AI in decision-making. Likewise, banks are exploring bespoke AI infrastructure and partnerships – for instance, investing in dedicated AI hardware and models – to reduce reliance on Big Tech providers and gain a competitive edge. These aggressive moves by both fintechs and traditional banks show a broad recognition that AI will separate the leaders from the laggards in financial services.

For senior banking executives, these developments sharpen a strategic question: should they build AI capabilities internally or acquire them? Stripe’s multi-billion purchase and Goldman’s in-house initiatives illustrate two paths to the same goal of AI leadership. Early movers are aiming to gain outsized efficiency and revenue advantages by infusing AI deeply into trading, lending, and operations. In an AI-charged market, waiting on the sidelines could mean ceding ground to more ambitious competitors – a risk no bank or insurer can afford in the long run.

This week also marked a breakthrough in autonomous finance as a concept becomes reality. FinTech bank Mercury introduced new 'Agent Cards' that allow businesses to issue virtual payment cards to AI-powered systems, enabling those bots to initiate and complete purchases within predetermined spending limits and budgets . While in the past AI in finance has mostly been confined to back-office analytics or customer chatbots, Mercury’s move – along with a similar service being piloted by corporate card provider Ramp – shows that AI agents are now being entrusted with real financial transactions. It’s a small but significant step toward a future where software agents handle routine financial operations, from procurement to treasury management, with minimal human input.

Established players are taking notice. At a major industry event in Europe just weeks ago, Mastercard, ING and Worldline demonstrated the continent’s first end-to-end payment made by an AI “digital assistant,” which was able to purchase a product online with only a single human approval step . That proof-of-concept underscored that agent-driven transactions can work on existing payment rails. Now, Mercury’s launch is translating that concept into a live product for everyday business use. The implication for banks and payment providers is clear – they will need to accelerate their own autonomous finance experiments or risk being left behind if customers come to expect AI-enabled, hands-free financial services.

Early adopters of autonomous finance are already reporting dramatic efficiency gains. For instance, London-based insurer Hiscox found that using an AI underwriting agent slashed the time required to generate complex policy quotes from three days to about three minutes – a 99% reduction in turnaround ([1]). Such breathtaking improvements in speed and productivity highlight AI’s potential to reinvent routine processes in banking and insurance. At the same time, they reinforce why strong oversight is critical: as more decisions and transactions move to AI, human roles will shift toward strategic supervision, and firms must ensure their AI 'colleagues' act within well-defined risk limits.

A stunning market event underscored the need for caution amid the AI gold rush. In July, highly leveraged bets on artificial intelligence turned sour for famed trading firm Jane Street, which reportedly lost around $15 billion in a single month . The loss stemmed from the meltdown of an AI-focused hedge fund – run by a former OpenAI researcher – that imploded during a sharp selloff in tech stocks, forcing a fire-sale of its holdings and leaving Jane Street with its largest-ever drawdown . This episode has sent shockwaves through the industry, showing how swiftly a cascade of correlated AI-era trades can unwind, and raising fears about hidden concentrations of risk in the financial system.

Banks are also grappling with novel forms of fraud and market abuse enabled by AI. Regulators and law enforcement warn that criminals are exploiting generative AI to produce deepfake videos, synthetic identities, and cloned voices that make investment scams and social engineering more convincing than ever. This surge in AI-driven fraud is putting pressure on financial institutions to strengthen their defenses – from better identity verification to AI-based anomaly detection – as traditional controls may struggle to catch advanced forgeries.

Meanwhile, a significant trust gap threatens to limit the usage of AI in client-facing roles. A new Gallup survey found that about one in five Americans who sought financial advice over the past year have experimented with AI for guidance, but only roughly 30% had even “some” confidence in the answers – and a mere 3% had a high level of trust . In contrast, over 80% of respondents expressed confidence in human financial advisors. This stark disparity suggests that while consumers are willing to try AI tools, they still put far more faith in human expertise. Financial firms will need to focus on improving AI transparency, explainability, and human oversight if they hope to convert early curiosity into lasting adoption.

Global regulators are quickly moving to tighten oversight of AI in finance, aiming to prevent the next crisis before it happens. In Singapore, the Monetary Authority (MAS) used a parliamentary hearing this month to declare that its forthcoming AI governance guidelines will explicitly cover 'agentic' AI – autonomous systems that can make decisions or transactions – used by financial institutions ([1]). This makes MAS the first major regulator to extend formal compliance requirements to AI decision-making agents, well ahead of the European Union’s timeline for addressing such technologies. The clear signal is that any use of advanced AI in banking must abide by rigorous risk management and board oversight, just like traditional models.

In the EU, the long-anticipated Artificial Intelligence Act has now moved from theory to reality, bringing hefty new compliance obligations for banks and fintechs. As of August 2, “high-risk” AI systems deployed in credit scoring, fraud detection, trading algorithms, insurance underwriting and other sensitive areas are subject to strict requirements for risk assessments, transparency, human oversight and data governance ([2]). Penalties for breaching these rules are draconian: fines can reach €35 million or 7% of global annual turnover in the worst cases, nearly double the maximum under GDPR ([3]) ([4]). European supervisors are also focusing on immediate threats: the EU’s systemic risk watchdog recently warned that cutting-edge “frontier” AI models could pose severe cyber risks to the financial system if banks do not harden their defenses, prompting large firms to report their mitigation plans by October.

U.S. regulators, for their part, have so far held off on AI-specific rules, but they are not standing idle. Bank examiners at the Federal Reserve and OCC have begun scrutinizing how banks use AI in high-risk activities like lending, fraud monitoring and compliance, even suggesting 'kill switches' for algorithms that behave unpredictably ([5]). And in a notable move, JPMorgan Chase CEO Jamie Dimon is spearheading a cross-industry coalition of some 40 major companies to identify AI risks and develop shared safeguards in collaboration with the government ([6]) ([7]). The upshot for financial institutions is a regulatory environment that is rapidly evolving: those that proactively invest in AI governance, explainability and robust controls will be far better positioned than those forced into reactive compliance under pressure.