Finance at an AI Crossroads: Bold Bets, New Rules & Big Warnings

Banks across Europe just entered a new era of AI governance. On August 2, the EU’s landmark Artificial Intelligence Act moved from theory to reality, meaning “high-risk” systems – such as AI models used in credit scoring, fraud detection, anti-money-laundering (AML) monitoring, or insurance pricing – are now subject to rigorous compliance requirements ([1]). Financial firms must conduct detailed risk assessments, ensure human oversight of AI-driven decisions, and maintain auditable records for these systems. The stakes are high: violations of the AI Act can trigger fines as steep as €35 million or 7% of global annual turnover in worst cases ([2]), nearly double the maximum penalty under GDPR.

For banks and insurers, the new rules create an additional compliance layer on top of existing regulations like data protection and operational resilience. The immediate priority is to get “conformity assessments” and monitoring in place for any AI used in sensitive areas like lending or trading. This push is not limited to the EU – thanks to a “Brussels Effect,” Europe’s strict approach to AI is already influencing corporate governance globally ([3]). Proactive firms that align early with the AI Act have been found to demonstrate stronger risk management and even enjoy greater investor confidence ([4]). In other words, regulators see responsible AI as integral to financial stability and are nudging the industry worldwide to adopt these practices.

European authorities aren’t stopping at abstract principles – they are also zeroing in on immediate threats. The European Central Bank’s banking supervisor, Claudia Buch, has ordered the bloc’s largest banks to spell out by October how they will contain risks from cutting-edge AI models ([5]). One fear: advanced “frontier” AI like Anthropic’s latest systems could identify software vulnerabilities and generate exploits far faster than banks can patch them ([6]). Europe’s top systemic risk watchdog went so far as to label such AI-enabled cyber threats a potential systemic risk to the financial system if defenses aren’t strengthened ([7]). The message for executives is clear – as you adopt AI, regulators expect equally forceful action to manage its risks.

One of the world’s preeminent investment banks is moving AI from the lab to the core of its business. Goldman Sachs Asset Management has launched “AlphaAI,” a new artificial intelligence platform designed to drive investment decisions across both public markets and private equity ([1]). Long-time Goldman executive Lou D’Ambrosio was appointed as the firm’s first-ever Chairman of AI for Asset Management to lead the initiative ([2]) – a clear sign that Goldman is treating AI not just as an investment fad but as fundamental to how it competes. “We believe AI is both reshaping industries and acting as a force multiplier in how we invest,” said Goldman’s asset management chief, underlining that AI will be deeply embedded in identifying opportunities and managing portfolios ([3]).

Goldman’s move is part of a broader trend: large banks and asset managers are racing to develop proprietary AI capabilities enterprise-wide ([4]). Banks like Citi have rolled out internal “agentic AI” platforms that let employees build and deploy AI agents for tasks ranging from customer service to compliance, all within a secure, bank-controlled environment ([5]). JPMorgan, meanwhile, says it plans to deploy AI agents that can operate autonomously for extended periods once they clear internal governance hurdles ([6]). Rather than relying solely on off-the-shelf AI tools, firms are leveraging their vast troves of data and domain expertise to train custom models – from large language models to real-time fraud detectors – aiming for an edge in everything from trading strategies to customer engagement ([7]).

Scaling AI in finance is also prompting reorganization and re-skilling. JPMorgan’s leadership acknowledges that AI will automate some jobs, but instead of broad layoffs, the bank has “huge” plans to retrain and redeploy staff impacted by AI-driven changes ([8]). HSBC’s COO similarly committed to upskill all 200,000 employees as AI transforms the bank’s operations ([9]) – even as rival Standard Chartered announced it will cut about 15% of its support roles by 2030 in an AI-driven efficiency push ([10]). This divergence in strategy highlights a key decision point for executives: whether to view AI primarily as a cost-cutting automation tool or as a growth driver that amplifies human productivity ([11]). The consensus forming among leading firms is that the real prize lies in using AI to augment human capabilities, gain competitive advantage, and unlock new revenue streams – not just to trim headcount.

As AI becomes mission-critical, some banks are taking an unexpected route – investing in the infrastructure behind AI to reduce reliance on Big Tech. In a significant departure from the usual cloud-first approach, JPMorgan Chase has partnered with Silicon Valley AI chipmaker SambaNova, which just raised $1 billion at an $11 billion valuation ([1]). JPMorgan will be one of the first to deploy SambaNova’s next-generation “SN50” systems for its own AI needs, running large machine learning models on its premises behind the bank’s firewall ([2]). This tie-up “sends a message to the banking industry that it’s time not to completely depend on cloud services” and to embrace a more “heterogeneous” computing strategy ([3]).

Controlling their own AI hardware offers financial institutions two advantages: better data control and potential cost efficiency. Banks handle sensitive customer and trading data subject to strict regulations, and keeping AI processing in-house can reduce security and privacy risks. Cost is another factor – training and deploying advanced AI is expensive at the scale of big banks, contributing to an estimated $1 trillion collective AI infrastructure spend by the top five tech firms in 2025–2026 ([4]). If those massive investments don’t yield commensurate returns, a cloud-driven AI boom could quickly turn into a bust ([5]). By developing private AI clouds and teaming up with specialized chip providers, banks like JPMorgan aim to ensure they have the capacity and flexibility to run AI models efficiently, while avoiding overdependence on external cloud vendors.

This trend could spark a new wave of competition (and partnership) between banks and technology giants. Cloud providers may find themselves pressured to offer more flexible solutions – for instance, delivering enterprise AI systems that can run on a client’s own hardware – as big banks demonstrate willingness to invest in bespoke tech stacks. For financial leaders, the takeaway is that building internal AI infrastructure, once seen as costly and impractical, is increasingly on the table when the alternatives are multi-year cloud contracts with uncertain ROI and vendor risk. The largest banks are effectively hedging their bets: doubling down on AI, but on their own terms.

Amid the gold rush for AI, global financial authorities are injecting notes of caution. The Bank for International Settlements (BIS) – often called the central banks’ central bank – warned that the current wave of AI investment bears resemblance to history’s great investment manias ([1]). In its latest analysis, the BIS pointed out that the world’s five largest tech firms alone may pour over $1 trillion into AI-related capital expenditures from 2025 through 2026 ([2]). The concern is that wildly optimistic bets on AI could overshoot reality; if returns disappoint, a rapid pullback in funding could turn today’s euphoria into a protracted bust ([3]) – with potential spillover effects for financial markets and lenders financing the tech sector.

Companies at the center of the AI boom are themselves acknowledging risks. In a recent SEC filing, cloud giant Oracle – which has heavily invested in building data centers for AI clients – explicitly flagged the danger of customers failing to pay their bills or renew contracts ([4]). While not naming names, the disclosure is widely seen as referring to OpenAI, which has a massive multi-billion-dollar cloud deal with Oracle. If such AI clients can’t monetize their products quickly enough, providers could be left holding unused capacity – a scenario with echoes of past over-investment crises in tech. The takeaway: even as AI transforms finance, executives must stay vigilant about concentrations of tech exposure and perform robust due diligence on AI vendors’ resilience.

Financial regulators are also ramping up scrutiny of AI-driven trading and advice. In the U.S., lawmakers have pressed the Securities and Exchange Commission to crack down on “black box” robo-advisors and autonomous trading algorithms that manage consumer investments with little human intervention ([5]). The question regulators are grappling with is how to ensure these AI agents act in clients’ best interests – a standard no less stringent than for human advisers ([6]). For industry leaders, it’s a reminder that embracing AI in areas like wealth management and market trading comes with a need for transparency and accountability. Guardrails on AI models – from explainability standards to human override mechanisms – will be essential to maintain trust and compliance as algorithmic strategies become more prevalent.

Disruption from AI isn’t only coming from Big Tech or incumbents – fintech upstarts are leveraging AI to go straight after banks’ core businesses. In one of the largest recent fintech funding rounds, Berlin-based startup Taktile raised $110 million led by Goldman Sachs to scale its AI-driven decision platform for financial services ([1]). Taktile’s technology automates critical judgment calls in lending, insurance underwriting, fraud detection and Know-Your-Customer checks. The company boasts that its software can handle 95% of business loan underwriting decisions with no human input and reduce false fraud and AML alerts by 75% ([2]) – potentially enabling faster lending decisions with fewer compliance bottlenecks.

Such figures illustrate why investors are pouring capital into AI “decisioning” platforms. Taktile’s clients already include established players like U.K. digital bank Monzo and global insurer Allianz ([3]), showing that even incumbents are turning to specialist AI startups to upgrade their credit risk and compliance processes. In the insurance sector, larger vendors are also acquiring AI-native firms to boost innovation: for example, core systems provider Duck Creek recently bought out a startup to embed an “agentic” AI underwriting engine into its platform ([4]). These deals suggest that legacy technology vendors and big financial institutions alike recognize the need to modernize legacy systems and harness advanced AI – or risk being outpaced.

For C-suite leaders in finance, the message is clear. The competitive moat around traditional banking and insurance products is eroding as AI lowers barriers to entry for tech-savvy challengers. Fintechs are targeting high-margin activities (from loan approvals to wealth management) with automation and intelligence, often backed by incumbent investors who see the writing on the wall. To stay ahead, established banks and asset managers must decide where to partner with AI innovators and where to build their own capabilities. The next phase of fintech disruption will be defined by those who can integrate AI into their operating model quickly and responsibly – achieving both efficiency gains and new revenue opportunities without compromising on risk management or customer trust.