News Summary
In September 2026, the European Central Bank launched Project Pontes, an operational initiative aimed at connecting distributed ledger technology with the Eurosystem TARGET settlement infrastructure.
Following remarks by ECB Executive Board member Isabel Schnabel at Jackson Hole on August 28, Project Pontes represents a near-term step toward Project Appia, the ECB longer-term initiative for developing a wholesale settlement framework. Meanwhile, the Bank for International Settlements and seven central banks delivered the prototype for Project Agorá, validating the feasibility of atomic settlement combining tokenized commercial bank deposits with tokenized central bank money.
These developments, alongside the Tokenised SEPA proposal from the Bank of Italy, signal a structural shift toward programmable and tokenized settlement architectures alongside existing account-based systems.
Settlement Architecture
Programmable settlement is bridging tokenized assets and central bank money.
As Isabel Schnabel has pointed out, the financial sector faces a growing infrastructure gap: tokenized securities and corporate bonds can be issued and traded on distributed ledgers, while cash settlement still largely relies on traditional off-chain payment systems such as TARGET.
This separation between tokenized assets and conventional settlement money can create operational frictions and additional settlement risks. Projects like Pontes and Agorá address this by bringing key asset classes onto shared DLT networks.
- Tokenized commercial bank deposits: digital representations of customer deposits issued by regulated commercial banks.
- Tokenized financial assets: securities and bonds that can execute directly on-chain.
- Tokenized central bank reserves: risk-free central bank money for final settlement.
Autonomous Execution
AI agents can turn programmable settlement into autonomous financial execution.
Traditional payment and settlement processes often include intermediary checks and processing intervals that leave time for reconciliation, screening, or manual intervention. Modern programmable networks change that model through conditional atomic settlement, where multiple transaction legs are completed simultaneously on an all-or-nothing basis.
Execution no longer has to rely on manual, multi-stage approval processes. Automated AI models can take over routine cash-flow management and execute market transactions based on preset parameters. As AI agents issue binding payment instructions directly to networks such as Pontes or Agorá, financial execution shifts from a manually driven workflow into autonomous, code-based execution.
Regulatory Accountability
Instant execution narrows the window for human intervention.
In traditional financial systems, processing delays provide an operational buffer that allows risk managers to review and halt suspicious or erroneous transactions. Under an atomic settlement model, once preset parameters are met, transaction legs are executed immediately and conditionally.
If an AI model experiences model drift, erroneous instructions may automatically trigger execution. That removes the manual intervention step found in traditional settlement processes and can force institutions into complex post-settlement reversal operations.
Autonomous execution does not displace existing regulatory accountability obligations. Regulated firms remain fully responsible for the systems they deploy. Under senior management accountability regimes such as the UK FCA SM&CR, designated individuals must demonstrate that they have taken reasonable steps to control, constrain, and monitor autonomous systems, even when no human intervenes at the exact moment of execution.
Looking Ahead
Execution provenance is becoming a compliance requirement.
The future of programmable finance will depend not only on tokenization technology and AI models, but on how these systems integrate into existing legal and regulatory frameworks. Central banks and financial regulators have made clear that autonomous execution does not operate in a regulatory vacuum.
Traceability of execution will emerge as a core compliance requirement. As financial institutions combine autonomous AI logic with atomic settlement infrastructure, they will need ways to track and verify AI decision parameters, system prompts, execution boundaries, and control conditions.
Industry standards for auditing algorithms on programmable infrastructure are likely to follow. Macro-level digital finance frameworks can establish general benchmarks, but operational concepts such as reasonable measures and verifiable execution in high-speed tokenized environments will require more detailed standards for autonomous market activity.
