When Chinese AI lab Deepseek released a state-of-the-art large language model for free, it marked a turning point in enterprise AI strategy — especially for financial institutions. The move underscored a broader shift: the emergence of open-source AI as a serious contender to proprietary offerings.

While open-source tools have long been foundational to developer ecosystems, their influence rarely extended into core infrastructure. That changed when companies like Meta Platforms Inc. META began publishing foundational models freely, echoing Google's Android strategy — build the ecosystem now, monetize later.

The trend is accelerating. Deepseek's release made waves for several reasons:

• State-of-the-art models are now widely available at no cost.
• Training these models has become more economically feasible.
• Open-source distribution can accelerate adoption — Deepseek's app reportedly reached the top of mobile app store rankings within days of launch.

These developments open the door for enterprises — including banks — to stop merely consuming AI and start building it.

Banking's Quiet AI Revolution

According to a 2024 Citibank report, up to 54% of banking roles could be impacted by AI. The sector's heavy reliance on manual workflows, compliance-heavy documentation and repetitive research makes it ripe for disruption.

So far, most large U.S. banks have opted to explore generative AI via partnerships rather than by developing their own infrastructure:

• Morgan Stanley MS has deployed tools for automating meeting notes and summarizing research, aimed at enhancing financial advisor productivity.
• JPMorgan Chase & Co. JPM partnered with OpenAI to launch internal assistants for email and document drafting, all integrated into its secure Omni platform.
• Goldman Sachs Group Inc. GS built an internal hub to access specialized AI tools, with a focus on developer productivity and task-specific model optimization.
• Bank of America Corp. BAC and Wells Fargo & Co. WFC are focused on customer-facing AI, deploying chatbots in mobile banking apps to answer user queries and handle support.

Why Banks May Need to Build Their Own AI

With open-source AI infrastructure advancing rapidly, banks now have compelling reasons to develop internal models:

🔐 Enhanced Data Security

By building proprietary LLMs, banks can maintain end-to-end control over sensitive data, reduce reliance on external vendors, and ensure compliance with strict regulatory requirements.

📊 Unlocking Proprietary Data Value

Banks are sitting on massive pools of transaction and behavioral data that could power competitive AI models. Training open-source models in-house offers a cost-effective way to turn this underutilized data into differentiated digital services.

💸 Monetization Potential

Custom bank-built AI can serve internal teams or be licensed externally. As demand rises for financial data integration, models trained on proprietary datasets could enable entirely new product lines or revenue channels.

The Strategic Fork in the Road

Banks are now faced with a strategic decision: continue outsourcing AI capabilities to tech firms — or invest in building their own. With massive IT budgets and uniquely rich data troves, financial institutions are well-positioned to become creators of infrastructure, not just users.

The road ahead is clear. As open-source AI evolves, banking giants that invest in their own models today may define the next era of financial services.

Read Next:
• U.S. Bancorp Beats Q1 Estimates, CEO Says ‘Revenues Outpaced Expenses’

Photo: Shutterstock