AUDIO INTEGRITY & VOICE PROTECTION
Detect manipulated audio. Protect authentic voices.
FakeXpose shows where audio may have been manipulated, second by second. For high-stakes decisions, forensic experts review the evidence and deliver a documented finding. Protect the recordings you control with an inaudible watermark designed to verify source provenance.
- Authentic
- Manipulated
- Escalate
Built for insurance, legal, financial-services, media and public-sector teams that need decisions they can explain and defend.
The stakes
Why hearing a voice is no longer enough
Fraud risk
$893M+
in AI-related fraud losses reported to the FBI Internet Crime Complaint Center in 2025.
Source: FBI IC3, 2025 Internet Crime Report
Trust risk
100journalists
targeted with cloned or manipulated audio across 27 countries between December 2023 and December 2025.
Source: Reporters Without Borders
Decision speed
Seconds to triage. Days to document.
Automated analysis identifies suspicious segments in seconds. A documented expert finding is delivered in 3–5 business days.
Protect before abuse
Watermark authorized audio before it is published, shared or used in operational workflows, then verify its origin if it later appears in a disputed file.
A convincing voice is no longer evidence of identity. Verify the recording, protect the source, and document the conclusion.
Who it's for
Built for teams that must verify, protect and defend
Detect suspicious audio, protect approved recordings, and obtain expert evidence when the decision matters.
Building a voice or media product? Add manipulation detection, source-protection watermarking, provenance verification and expert-review escalation to your workflow with the FakeXpose API. API overview
More industries
One platform, many audiences
Anyone who publishes, receives, stores or relies on audio and media can use FakeXpose to establish what is original.
The problem
A convincing voice is no longer proof of identity or origin.
Fakes are cheap to make
Voice-cloning tools can imitate a person from a short recording. A call, voicemail, statement or video soundtrack can be fabricated or partly edited without leaving anything a listener can reliably hear.
One verdict is not enough
A single “real” or “fake” label does not show where a recording changed, how confident the system is, or whether the conclusion can survive a challenge.
Mistakes are expensive
Pay a fraudulent claim, release a wire, accept a doctored exhibit or publish a false quote, and the financial, legal and reputational damage can be difficult to undo.
Detection starts after exposure
Once an authentic voice recording is public, it may be copied, edited or used as input to a cloning system. Protecting authorized source audio before distribution adds a layer of traceability.
Detection vs. source protection
Detection asks
“Was this file manipulated?”
Source protection asks
“Was this file derived from an approved recording, and can it still be linked to the original source?”
Watermarking and deepfake detection solve different but complementary problems. A valid source marker is a positive signal about origin. A missing marker does not prove a file is false or manipulated.
Can you hear the difference?
Two clips of the same sentence — one authentic, one synthetic.
Four capabilities, one workflow
Four capabilities. One audio-integrity workflow.
Detect suspicious audio, protect source recordings, and get expert support when a decision must be defended.
- Protect source audio
- Verify provenance
SourceGuard Watermark
Embed an inaudible, source-linked watermark into authorized audio, then check whether a questioned copy derives from the protected source.
Explore SourceGuard- Detect manipulation
Verify Now
Upload a recording and receive a second-by-second timeline showing suspicious segments and confidence signals. Use it for fast triage, investigation intake and routing.
Explore Verify Now- Escalate to expert review
Expert in the Loop
Forensic analysts review the relevant evidence, detection outputs, source and provenance signals, and case context. Receive a documented report from a named analyst for disputes, investigations and legal review.
See forensic expert review
APIs
Use verification, watermark embed and detect, and escalation through one API and the same core pipelines as the dashboard.
View the APIVERIFY NOW
See where audio may have been manipulated.
Most tools return one verdict for an entire file. Verify Now analyzes a recording second by second, showing segments with indicators consistent with authentic, altered or synthetically generated speech, and the confidence associated with each finding. Review the timeline, share it with an investigator or editor, and escalate high-stakes cases to a forensic expert.
Upload a recording.
Submit an audio file through the dashboard or API.
Locate the issue.
Review segment-level signals, confidence scores and, where relevant, source and provenance information.
Take the next step.
Route low-risk cases, request human review, or compare the file against a protected source record.
- Second-by-second manipulation-risk timeline
- Segment-level confidence signals and evidence indicators
- Results in seconds for supported files
- Source-provenance signal when a protected original exists
- Escalation to forensic expert review from the same case workspace
- API access for high-volume or embedded workflows
Verify Now can be used with SourceGuard to assess both manipulation signals and whether a questioned file remains linked to an authorized recording.
FORENSIC EXPERT REVIEW
A named forensic expert behind the decision.
Automated analysis helps identify where to investigate. It does not replace professional judgment when the outcome affects a denied claim, payment release, public correction, regulatory inquiry, internal investigation or court matter. FakeXpose forensic analysts review the relevant audio, detection findings, provenance and watermark evidence where available, and case context to deliver a documented opinion you can act on and explain.
- Named forensic analyst accountable for the scope and findings
- Review of automated detection, source records and watermark/provenance evidence where available
- Segment-level analysis and documented methodology
- Report for investigations, claims files, disputes, records and legal-team review
- Optional forensic consulting to help you interpret results and plan next steps
- Standard report delivery in 3–5 business days
What you receive
- Case and file identification details
- Scope of the examination
- Segment-level timeline of the audio analyzed
- Detection findings and confidence context
- Watermark and provenance findings, if applicable
- Explanation of the method and its limitations
- Analyst name, role and credentials
- Conclusion framed to the evidence examined
Need help with a disputed recording, investigation, claims decision or litigation strategy? Our team can help define the examination scope, interpret findings, prepare documentation, and coordinate with your legal, fraud, claims or communications teams.
SOURCEGUARD WATERMARK
Protect source audio before it is shared.
Detection helps when suspicious audio has already appeared. SourceGuard helps protect authorized recordings before they are distributed or reused. FakeXpose embeds an inaudible, source-linked watermark into recordings you control. The protected audio stays usable in normal workflows while carrying a signal that can help verify provenance and identify links to the protected source.
Same listening experience. Added source protection and verification capability.
What SourceGuard can, and cannot, tell you
A valid source marker can provide positive evidence that a file is connected to a protected FakeXpose source. No marker found does not mean a recording is authentic or fake. The file may never have been protected, the signal may have been degraded, or the file may be unrelated to the protected source. Use Verify Now to assess manipulation indicators and Expert in the Loop for high-stakes decisions.
SourceGuard is a protective and forensic control, not an absolute guarantee. Results depend on the source audio, processing steps, file transformations and attack method. FakeXpose reports what the system detected and the confidence of that result.
Suggested uses
- Executive communications
- Spokesperson audio
- Official announcements
- Interviews
- Customer-service calls and prompts
- Licensed voice assets
- Broadcast masters
- Inaudible watermarking for supported formats
- API and dashboard workflows
- Protected-source record and file identity
Which product do I need?
Different questions, one platform.
| Verify Now | Expert in the Loop | SourceGuard Watermark | |
|---|---|---|---|
| Question | Is this audio manipulated? | Will this finding hold up? | Can I protect this recording, and check a suspect file against it? |
| You provide | A recording | A recording or flagged finding | Audio you own (protect) or a file to check (verify) |
| You get | Second-by-second timeline and confidence | A documented report from a named analyst | A protected file, or a source-relationship result |
| Speed | Seconds | 3–5 business days | — |
| Best for | Triage and screening | Disputes, investigations, legal review | Protection and provenance |
- Question
- Is this audio manipulated?
- You provide
- A recording
- You get
- Second-by-second timeline and confidence
- Speed
- Seconds
- Best for
- Triage and screening
- Question
- Will this finding hold up?
- You provide
- A recording or flagged finding
- You get
- A documented report from a named analyst
- Speed
- 3–5 business days
- Best for
- Disputes, investigations, legal review
- Question
- Can I protect this recording, and check a suspect file against it?
- You provide
- Audio you own (protect) or a file to check (verify)
- You get
- A protected file, or a source-relationship result
- Speed
- —
- Best for
- Protection and provenance
What our tools can and cannot do
Straight answers about the limits
Detection is probabilistic.
Scores show confidence, not certainty. That is why expert review exists.
Protection applies to audio protected before it is shared.
Recordings already public without a mark are not covered, and no protection is absolute.
Watermarks can be degraded.
Compression, editing and deliberate removal attempts can weaken a marker.
No marker does not mean fake, or authentic.
Use screening and expert review to decide.
Audio quality matters.
Noise, heavy compression and phone-line audio can affect results.
We are not a law firm.
Whether evidence is admitted is a decision for the court.
Privacy and security
Your audio stays your audio.
Recordings in a claim, a case file or an investigation are sensitive. Here is how we handle them.
- Role-based access and audit logs in the dashboard
Regulatory tailwinds
Built for a fast-moving standards landscape
NAIC Model Bulletin
Guidance to insurers on governing AI systems, including third-party vendor oversight.
Proposed FRE 707
Would apply reliability standards to machine-generated evidence. Published for comment in 2025–26 and sent back for further study in June 2026; not adopted.
OCC / FFIEC guidance
Supervisory expectations for documented AI-fraud governance and vendor oversight.
C2PA / ISO/IEC 22144
A content-provenance standard.
General information, not legal advice.
Start with one recording
Send us the recording you are unsure about.
Bring a real case. We run it through Verify Now, walk you through the timeline, and show how a flagged recording becomes a documented finding. Ready to protect your own audio? We will help you mark a first file and test verification.
- Choose a recording.
- Review the results together.
- Decide the next step: a full expert report, protecting your source audio, or the API.
Verify Now is the fast automated screen. Expert in the Loop adds a named forensic analyst and a documented report. SourceGuard protects recordings you control and lets you check later whether a file came from the protected source.
Verify Now returns results in seconds. Expert reports take 3–5 business days.
It is designed to be inaudible.
Media Shield is the company. FakeXpose is our flagship product — the audio detection, verification, and source-protection platform. As we build new tools for video and image authentication, they'll launch under the Media Shield name alongside FakeXpose.
Verify the recording. Protect the source. Document the conclusion.
Revealing Truth, Dispelling Myths