How does Chime use AI across its payment ecosystem?
In payments, Chime uses AI to score card payments, transfers and disputes for fraud, on infrastructure it owns. Its FY2025 10-K says Chime built its own machine learning platform for fraud, risk and underwriting decisions, with models retrained on cycles as short as three weeks and sub-second predictions fed by a real-time data stream. That stream, built on AWS and described in September 2023, ingests more than 1 million events per second, including card swipes and peer-to-peer payments. In August 2026 Chime's engineers reported moving four production models to transformer event-sequence models: a card-authorization model keeps the current amount and merchant name as real-time inputs, and the first-party fraud score, in production daily, improved recall for 99th-percentile-risk users by 20%. Chime says its AI-powered fraud detection and risk-scoring models helped cut fraud losses by 29% since 2022, and its AI disputes platform, introduced in 2024, cut average resolution time from about 42 minutes to under 20. Since November 2025 all transactions run on ChimeCore, its own processor and ledger, which replaced Galileo. Chime does not disclose fraud rates or authorization approval rates.[1][2][3][4]
Data as of Oct 19, 2026 · Chime profile updated Oct 16, 2026 · 12 sources cited on this page · 52 fintechs profiled on this site · Markdown
Where in the payment flow does Chime use AI?
One row per layer, with the latest dated description Chime has published. Results are Chime's own measurements.
| Layer | System | What the record shows | Date |
|---|---|---|---|
| Real-time event data for fraud | Streaming 2.0 (AWS Kinesis, Glue, DynamoDB, SageMaker) | Captures card swipe events, peer-to-peer payments, logins and enrollments; serves rule-based and ML fraud models with millisecond latency; more than 1 million events per second, end-to-end latency under 15 seconds, and model score calculation 1200% faster than the legacy system | September 19, 2023[2] |
| Model platform | In-house machine learning platform | Models for fraud, risk and underwriting decisions; auto-retrained and backtested, with retraining timelines as short as three weeks; sub-second latency predictions "critical to identifying and preventing fraud" | FY2025 10-K[1] |
| Model factory | MLKit | Configuration-driven framework built after Chime's ML footprint grew to over 40 models in production, from fraud detection to marketing; real-time models run on SageMaker endpoints | February 11, 2025[5] |
| Disputes and chargebacks | Risk decisioning platform | Uses first-party data including member income, spending, transaction and social graph data to help reduce costs of transaction-related disputes and chargebacks; also powers underwriting for Chime-branded liquidity products | FY2025 10-K[1] |
| Card authorization | Transformer event-sequence model | Keeps "the current transaction amount and merchant name" as real-time context for a card-auth model; the fraud model has about 10 million parameters and is served from the existing real-time scoring infrastructure | August 27, 2026[3] |
| First-party fraud | Transformer event-sequence model | Recall for 99th-percentile-risk users improved by 20%; the score runs daily in production and routes disputes into agent queues and prioritizes account reviews | August 27, 2026[3] |
| Merchant risk | Merchant Risk anomaly detection model | Adds 13% PR-AUC lift over its gradient-boosted-tree baseline | August 27, 2026[3] |
| Dispute handling | AI-powered disputes management platform | Introduced in 2024; average dispute resolution time fell from about 42 minutes to under 20; AI fraud detection and risk-scoring models helped reduce fraud losses by 29% since 2022, with final enforcement actions governed by rules and human oversight | January 14, 2026[4] |
| Dispute fraud (patent) | Patent application US20250061461A1 | Describes a fraud detection machine-learning model that generates a fraud prediction for a digital claim disputing a transaction, then authorizes or remediates the claim | Published February 20, 2025[6] |
| Stopping charges for members | Jade Financial Partner | In-app AI co-pilot that can stop future recurring charges and cancel supported subscriptions; available to some members | Help page retrieved September 23, 2026[7] |
Why does ChimeCore matter for Chime's AI in payments?
ChimeCore, launched in 2024, is Chime's proprietary payment processor and ledger. Per the FY2025 10-K, as of November 2025 it processes all payments, transfers, deposits, withdrawals and other transactions on the platform, and member transactions previously processed by Galileo Financial Technologies were moved to ChimeCore by the end of 2025.[1]
Chime ties the core to its AI work. Its January 2026 AI series says ChimeCore lets it embed AI "without relying on black-box vendors or legacy infrastructure", and CEO Chris Britt's 2025 shareholder letter says that with a member's permission Chime can "stop an unwanted charge before it hits" because ChimeCore runs from the ledger to the app.[8][9]
The same letter says transaction processing costs are already down an estimated 60% and build cycles fell from 12 weeks to as little as three days. Chime's September 8, 2026 agreement to buy Stride Bank says integrating ChimeCore with Stride's banking infrastructure "will unify data, decisioning, and reduce handoffs"; the deal had not closed and is expected to close in the first half of 2027, subject to OCC and Federal Reserve approval.[9][10]
Who does Chime work with on payments AI?
| Partner | What | Date |
|---|---|---|
| Amazon Web Services | Kinesis, Glue, DynamoDB and SageMaker behind the Streaming 2.0 fraud pipeline (2023); SageMaker pipelines and endpoints for MLKit models (2025); hosts Chime's platform (FY2025 10-K) | 2023–2026[2][5][1] |
| Anthropic (via Amazon Bedrock) | Claude powers Chime's AI call-summarization tool for support agents, in production since February 2024, per an AWS case study | 2024–2026[11] |
| The Bancorp Bank and Stride Bank | Chime develops risk and underwriting standards on its risk decisioning platform for review and adoption by its bank partners | FY2025 10-K[1] |
| Galileo Financial Technologies | Former third-party processor; member transactions moved to ChimeCore as of the end of 2025 | FY2025 10-K[1] |
| Stride Bank (pending acquisition) | Agreement to buy for $590 million in cash; Stride would become Chime Bank, N.A.; closing expected in the first half of 2027 | September 8, 2026[10] |
What does Chime say about AI-driven fraud risk, and what does it not disclose?
Chime's Q2 2026 10-Q says the increasing use of AI, "including generative and agentic AI, has accelerated the sophistication of fraudulent activity, including deepfakes, synthetic identity fraud, and automated attacks that may evade traditional detection tools", and that Chime has suffered and expects to continue to suffer losses from fraudulent activities.[12]
Transaction and risk losses were $103.3 million in Q2 2026 against $98.2 million in Q2 2025, and $192.2 million in the first half of 2026 against $207.4 million a year earlier, per the 10-Q. The line mainly covers losses on liquidity products and overdrawn accounts as well as transaction dispute losses, per the 10-K, so it is not a fraud-loss series.[12][1]
The August 2026 engineering post says the remaining task-specific transformer models are still to be shipped to real-time production, so the first-party fraud score is the only one described as live.[3]
What is not disclosed: card authorization approval rates, fraud losses in dollars or basis points, how many authorizations are scored by machine learning in real time, results of the dispute patent, and any agentic-commerce product for paying merchants through AI agents.[1][12]
What does the Chime profile record on this topic?
The rows below are the entries of the Chime AI strategy profile that bear on this page, as of Oct 16, 2026; the profile holds the full record of 13 dated moves, 8 use cases and 15 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| AI fraud detection and risk scoring | AI-powered fraud detection and risk-scoring models that Chime says helped reduce fraud losses by 29% since 2022, with final enforcement actions governed by rules and human oversight. | In production | [4] |
| AI disputes management platform | Introduced in 2024; average dispute resolution time fell from about 42 minutes to under 20, and mail processing and document reviews are 70% faster. | In production | [4] |
| Transformer event-sequence risk models | Transformers consume raw member event sequences for lending default, first-party fraud, merchant risk and churn models, with a 13–35% relative PR-AUC lift; the first-party fraud score runs daily in production, and the remaining task-specific models are still being moved to real-time production. | Rolling out | [3] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Sep 8, 2026 | Agreement to acquire Stride Bank | Chime agrees to buy its partner bank Stride Bank, N.A. for $590 million in cash; Stride would become Chime Bank, N.A., subject to OCC and Federal Reserve approval, with closing expected in the first half of 2027. | [10] |
| Aug 27, 2026 | Transformer risk models disclosed | Chime's engineers report replacing feature engineering in four production models (lending default, first-party fraud, merchant risk, direct-depositor churn) with transformer event-sequence models, capturing 12% more defaulters in the top decile for lending. | [3] |
| Jan 14, 2026 | Member-experience AI results published | Chime reports that generative AI voice and chatbots introduced in 2025 now resolve about 66% of self-service calls and about 75% of chats, that its AI disputes platform (2024) cut average resolution time from about 42 minutes to under 20, and that AI fraud and risk-scoring models helped cut fraud losses 29% since 2022. | [4] |
Does Chime use AI to detect fraud?
Yes. Chime's FY2025 10-K says it built its own machine learning platform for fraud, risk and underwriting decisions, with sub-second predictions from real-time data. Chime said in January 2026 that AI-powered fraud detection and risk-scoring models helped reduce fraud losses by 29% since 2022, with final enforcement actions governed by rules and human oversight.
How does Chime score card transactions in real time?
Its Streaming 2.0 pipeline on AWS, described in September 2023, captures card swipes and peer-to-peer payments and serves fraud models with millisecond latency at more than 1 million events per second. In August 2026 Chime said a transformer-based card-authorization model uses the current transaction amount and merchant name as real-time context.
What is ChimeCore?
ChimeCore is Chime's proprietary payment processor and ledger, launched in 2024. As of November 2025 it processes all payments, transfers, deposits and withdrawals on Chime's platform, replacing the third-party processor Galileo. Chime says it has cut transaction processing costs by an estimated 60% and lets it embed AI without relying on outside vendors.
How does Chime use AI in payment disputes?
Chime's AI-powered disputes management platform, introduced in 2024, cut average dispute resolution time from about 42 minutes to under 20, Chime said in January 2026. Its transformer first-party fraud score, which Chime said in August 2026 runs daily in production, routes disputes into agent queues. Chime also has a published patent application for a dispute fraud-prediction model.
Which technology partners does Chime use for payments AI?
Amazon Web Services runs the streaming, feature-store and SageMaker components behind Chime's real-time fraud models, and Anthropic's Claude via Amazon Bedrock powers its support call summaries. Chime sets risk and underwriting standards for its partner banks, The Bancorp Bank and Stride Bank, and agreed in September 2026 to buy Stride.
- Chime Financial, Inc. Annual Report on Form 10-K for the fiscal year ended December 31, 2025
- How Chime Financial uses AWS to build a serverless stream analytics platform and defeat fraudsters
- From Tabular to Transformers: Event Sequence Models for Financial Intelligence
- How Chime Is Using AI to Redefine Member Experience
- From redundancy to efficiency: Transforming machine learning operations at Chime
- US20250061461A1: Generating a fraud prediction utilizing a fraud-prediction machine-learning model (Chime Financial, Inc.)
- What is Jade Financial Partner? (help centre, retrieved 2026-09-23)
- How Chime Is Using AI to Rewire Financial Services
- 2025 Letter to Shareholders
- Chime Announces Agreement to Acquire Stride Bank
- Chime Financial improves member experience and saves 250,000+ hours annually with AI-powered call summaries using Amazon Bedrock
- Chime Financial, Inc. Form 10-Q for the quarter ended June 30, 2026
When Chime moves on AI, you'll read it here first.
when a bank, fintech or regulator moves on AI, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning