Global banks are quietly handing over more of their currency-trading decisions to machines, and the latest sign of that shift comes from Ant International, which just upgraded its Falcon forecasting system and locked in deals with five of the world’s biggest lenders. The move signals that AI foreign exchange forecasting is no longer a lab experiment for banks like Citi, HSBC, Deutsche Bank, Standard Chartered, and Barclays — it’s becoming part of how they actually manage money.
Summary
Key takeaways
- Ant International upgraded its Falcon FX forecasting model to version 2.0, a transformer-based system with nearly two billion parameters.
- Citi, HSBC, Deutsche Bank, Standard Chartered, and Barclays have all partnered with Ant to use Falcon in foreign currency trading.
- Standard Chartered says Falcon delivers over 90% forecasting accuracy and cuts liquidity management costs by 50%.
- Global FX turnover hit $9.5 trillion a day in April 2025, according to the Bank for International Settlements, partly due to tariff-driven hedging.
- The Financial Stability Board has flagged concerns that widespread reliance on similar AI models could create systemic risk across financial markets.
Ant International Launches Falcon 2.0 for FX Forecasting
Ant International confirmed on Thursday that it has pushed its currency-forecasting engine, Falcon, to version 2.0, marking a deeper push into AI foreign exchange forecasting as banks look for cheaper ways to manage currency risk. Kelvin Li, Ant International’s general manager of platform technology, framed the update as a direct answer to the limits of general-purpose AI in finance. “Precise forecasting can slash foreign exchange hedging and allocation costs by over 60%,” Li said, adding that general-purpose large models have “yet to achieve a universal breakthrough in the financial sector.”
Technical Features of Falcon AI
Falcon runs on a transformer-based architecture with nearly two billion parameters, putting it in a different category from the broad, general-use language models most people associate with AI. Rather than trying to do everything, Falcon is built specifically to predict cash flow and currency exposure — the kind of narrow, high-stakes forecasting that banks depend on for hedging decisions.
Open Access via ArXiv and GitHub
Unlike most proprietary bank forecasting tools, Ant documented the Falcon 2.0 release this month in a technical report on arXiv and published its code on GitHub. That level of openness lets outside researchers and rival institutions scrutinize the model’s methodology, a notable departure from how banks typically guard their internal risk systems.
Global Banking Partnerships and Applications
Falcon’s reach now extends across five major banks, each testing or deploying the model for different slices of currency risk management. These bank AI partnerships didn’t appear overnight — several banks had already been running pilots with Ant for more than a year before this week’s announcement formalized the relationships.
Major Banks Adopting Falcon
Li named Citi, HSBC, Deutsche Bank, Standard Chartered, and Barclays as Falcon’s banking partners. The lineup matters beyond traditional FX trading, since several of these institutions are simultaneously building tokenized deposit and digital payment infrastructure aimed at enabling round-the-clock cross-border transfers.
Pilot Results and Use Cases
Citi launched its Falcon pilot in July 2025 alongside its Fixed FX Rates product, which lets online retailers lock in exchange rates across more than 70 currencies. Citi reported that pairing the two tools helped an airline client cut hedging costs by roughly 30%. Standard Chartered took a different route, combining Falcon with its Aggregated Liquidity Engine, known as SCALE. The bank says that setup forecasts Ant’s currency exposures with over 90% accuracy while reducing liquidity management costs by 50%.
Tokenized Currency and Cross-Border Payments
HSBC’s collaboration with Ant goes further than FX forecasting alone. The two built a Tokenized Deposit Service and used it in 2025 to process a cross-border payment on the ISO 20022 messaging standard. That overlap is worth watching — AI forecasting and tokenized settlement tackle opposite ends of the same problem, predicting where liquidity will be needed and then moving it there faster.
Market Context and Financial Implications
Currency risk has become impossible for multinational firms to shrug off, which is exactly why FX hedging cost reduction tools like Falcon are landing at the right moment. According to the latest data, daily average foreign exchange trading reached $9.5 trillion in April 2025, representing a 27% increase compared to the same period three years prior, as reported by the Bank for International Settlements. The BIS tied part of that surge to companies scrambling to hedge dollar exposure after US tariff announcements rattled currency markets.
Higher interest rates since 2022 have made some hedges more expensive too, forcing investors to weigh the cost of protection against the risk of leaving exposure unhedged. Against that backdrop, a system that promises to cut hedging and allocation costs by more than 60% has obvious appeal — not just for Ant, but for any bank trying to keep clients from bleeding money on currency swings.
Regulatory and Systemic Risk Considerations
Here’s the catch: when several of the world’s largest banks lean on the same AI model, a single flawed forecast could ripple across the entire system instead of staying contained to one institution. The Financial Stability Board, in research compiled by the BIS and released in June 2025, cautioned that institutions in the financial sector depending on a limited number of AI solution providers — or leveraging comparable models developed using comparable datasets — risk creating system-wide vulnerabilities.
The concern isn’t just that one model might misfire on a single trade. If multiple large banks are all reading from the same signals, errors or embedded biases could push them toward correlated decisions at exactly the moment markets are already under stress. Falcon’s rapid adoption cuts both ways: wider use could make FX hedging and liquidity management genuinely cheaper, while simultaneously building a shared technological dependency that regulators are now watching closely.
Ant International’s Growth and Strategic Investments
Ant isn’t slowing down. In an equity financing round completed the previous month, the company secured $1.2 billion, resulting in fresh capital to push its financial AI deeper into global banking. That war chest suggests Falcon’s rollout across Citi, HSBC, Deutsche Bank, Standard Chartered, and Barclays is just the opening phase of a broader bet that AI foreign exchange forecasting will become standard infrastructure for how banks price and hedge currency risk — even as regulators sort out what happens if too many of them end up trusting the same machine.
FAQ
What is Ant International’s Falcon model?
Falcon is a transformer-based AI model with nearly two billion parameters that forecasts cash flow and FX exposure, achieving over 90% accuracy and processing over 60% of Ant’s FX conversions.
Which major banks are using Falcon for foreign currency trading?
Six major banks, including Citi, HSBC, Deutsche Bank, Standard Chartered, and Barclays, have partnered with Ant International to use Falcon in their foreign currency trading.
How much can Falcon reduce foreign exchange hedging costs?
Falcon can reduce foreign exchange hedging and allocation costs by over 60%, with Citi reporting approximately 30% savings in a pilot and Standard Chartered cutting liquidity management costs by 50%.
Are there any regulatory concerns about using AI models like Falcon in financial markets?
Yes, the Financial Stability Board has warned that reliance on similar AI models by many institutions could create systemic risks through correlated market actions during periods of stress.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
