Sound Alibi

Verify the story. Not just the identity.

Inspiration

What if the biggest weakness in banking security isn't a stolen password, but a trusted customer being tricked into sending their own money?

Traditional fraud prevention focuses on verifying who is making a transaction. But modern scammers don't always need to hack accounts. They manipulate people through fear, urgency, affection, and convincing stories until victims willingly authorize payments.

A grandmother believes her grandchild needs emergency money. A student receives a convincing tuition payment request. Someone gets a call claiming to be from their credit union, asking them to transfer money to protect their account.

In each case, the transaction may be authorized, but the decision is being manipulated.

That realization inspired Sound Alibi, an AI-assisted scam prevention experience designed for Sound Credit Union.

Our guiding question became:

What if banking security could help verify the story behind a payment, not just the identity of the person sending it?

What It Does

Sound Alibi introduces an additional layer of protection between a suspicious payment request and a potentially irreversible financial decision.

Rather than relying on generic fraud alerts, Sound Alibi helps members recognize manipulation, independently verify suspicious requests, and access assistance when they need it.

1. Explainable Scam Risk Detection

Sound Alibi evaluates transaction signals such as unusual payment amounts, new recipients, unfamiliar payment methods, and recent account changes.

Instead of presenting an unexplained warning, it shows why a transaction may deserve additional attention.

2. Personalized Alibi Missions

When a potentially suspicious transaction is identified, Sound Alibi can ask the member why they're sending money.

Based on the voluntarily provided context, the system recommends an independent verification action.

For example, if someone is sending $8,500 to help a grandchild supposedly facing an emergency, Sound Alibi can recommend calling the grandchild using a previously known phone number rather than trusting the caller's instructions.

The goal is not simply to tell members that something looks suspicious. It is to give them a practical way to find out.

3. Live Rescue

Scams often begin before a transaction exists.

Sound Alibi's Live Rescue experience gives members a way to seek help while receiving a suspicious call or message.

Members can identify warning signs such as requests for one-time passcodes, account credentials, urgent transfers, or remote device access.

The system provides immediate safety guidance and can create a case for employee review within the prototype.

4. Purpose Lock

Some scams involve changing payment instructions after trust has already been established.

Purpose Lock records payment details confirmed by the member and highlights changes before submission, helping members notice when the destination or purpose no longer matches their original understanding.

5. Recovery Autopilot and Evidence Vault

Not every scam is stopped in time.

For members who believe they've already sent money to a scammer, Sound Alibi provides guided next steps, supports voluntary evidence collection, and organizes information for follow-up.

It does not guarantee that transferred funds can be recovered.

6. Employee Investigation Dashboard

Sound Alibi also supports credit union employees through a dedicated investigation interface.

Employees can review flagged payments, Live Rescue requests, recovery cases, risk explanations, and case timelines.

This connects member-facing intervention with a structured employee workflow.

7. Fraud Intelligence

A fraud analytics dashboard helps illustrate suspicious transaction patterns, investigate individual payments, and examine the tradeoff between scam detection and unnecessary interruptions to legitimate members.

How We Built It

We designed Sound Alibi as a prototype of a combined member-facing and employee-facing fraud prevention platform.

The system brings together several components:

• Transaction risk analysis: Explainable rules evaluate available transaction and account activity signals.
• AI-assisted guidance: Context-aware assistance helps turn suspicious payment situations into actionable verification steps.
• Member experience: Interactive flows guide members through suspicious payments, urgent scam situations, and recovery.
• Employee experience: Investigation queues and case summaries help employees understand and respond to reported incidents.
• Fraud analytics: Structured transaction data supports risk analysis, investigation, and evaluation.

We worked with a synthetic financial activity dataset containing 7,471 events, 764 payments, 118 members, and 48 scam journeys.

The dataset represents approximately $340,546 in attempted payments, $213,608 sent, and $19,228 recovered.

These figures describe the synthetic scenarios used for the prototype, not money protected or recovered by Sound Alibi in production.

Challenges We Faced

Detecting manipulation instead of just unusual transactions

A large transfer isn't necessarily fraudulent. Someone purchasing a home may legitimately send a substantial wire, while a smaller payment could still be part of a scam.

We needed to consider both transaction signals and the story behind the payment without treating every unusual transaction as malicious.

Protecting members without creating unnecessary friction

Overly aggressive warnings can delay legitimate payments and frustrate members.

We focused on explainable alerts, independent verification, and workflows that allow legitimate members to continue through appropriate review.

Making scam prevention understandable

Fraud detection is only useful if people know what to do with the warning.

We designed guidance around specific actions rather than technical risk terminology, with a simplified experience for members who prefer clearer, step-by-step instructions.

Connecting the member and employee experiences

A warning alone doesn't solve the problem when someone needs immediate help.

We designed the member-facing rescue experience and employee investigation workflow to complement each other.

What We Learned

The most important lesson was that authentication and trust are not the same thing.

A member can pass every identity check and still be acting under a scammer's influence.

We also learned that effective fraud prevention requires more than a risk score. It needs clear explanations, independent verification, accessible guidance, and a practical path to human assistance.

Finally, we learned how important it is to balance security with member autonomy. A good intervention should help people make informed decisions, not simply create more barriers.

What's Next

Our next steps are to explore:

• Integration with eligible credit union payment and case-management workflows.
• More advanced scam-pattern analysis and evaluation using appropriately authorized data.
• Expanded accessibility and multilingual member guidance.
• Better measurement of false positives, intervention completion, and legitimate payment delays.
• A potential Subscription Shield feature that helps members review introductory offers, recurring billing terms, and potentially misleading subscription conditions before making a purchase.
• A controlled pilot to evaluate usability, employee workflows, and scam-prevention effectiveness.

Sound Alibi is currently a hackathon prototype, not a deployed banking security system. Its transaction outcomes and employee workflows are demonstrated using simulated data and interactions.

Why It Matters

Scammers are increasingly skilled at convincing people to participate in their own financial exploitation.

Banks already work hard to verify identities and detect suspicious activity. Sound Alibi explores what happens when we add another question:

Does the member have a trustworthy reason to send this money?

Because the future of scam prevention isn't just about protecting accounts.

It's about protecting the people behind them.

Sound Alibi — Verify the story. Not just the identity.

Built With

• ai
• api
• claude-api
• data-analytics
• fastapi
• fintech
• fraud-detection
• javascript
• python
• react
• rest
• tailwind-css
• typescript
• vite