DeepFakeCheck Blog

Review C2PA multi-asset part locations, coverage, hash references, and optional parts without confusing file integrity with event truth.

Check C2PA cloud-data required fields, hard-binding labels, update-manifest actions, and exact failure codes without treating structure as truth.

Validate C2PA actions fields, created/opened order, ingredient references, redactions, and exact failure codes without treating conformance as truth.

Check the single C2PA claim box, CBOR structure, required fields, signature URI, hashed references, and exact failure codes before judging provenance.

Check required and forbidden C2PA assertions by manifest type, record exact structural failure codes, and keep conformance separate from event truth.

Check each URI, algorithm, file count, and full-file hash in a C2PA collection while keeping integrity separate from claims about the files.

A C2PA 2.2 workflow for tracing hard bindings, hashing the asset, and recording match or mismatch without claiming the depicted event is true.

A practical C2PA 2.2 workflow for checking ZIP hashes, manifest placement, and signature results without claiming document truth.

Review C2PA signer and CA certificate revocation, OCSP evidence, timing, and status codes without mistaking credential trust for proof that media is true.

Learn what Well-Formed, Valid, and Trusted mean in C2PA, which checks each state requires, and why none proves that a depicted event is true.

Check each ingredient relationship, trace referenced manifests through the full lineage, and reconcile C2PA validation results without treating them as proof of an event.

A file-based workflow for resolving C2PA assertion URIs, recording missing or outside-manifest references, and checking hashes without claiming event truth.

Follow the C2PA 2.2 search order for embedded, HTTP Link, XMP, font, and sidecar manifests, then confirm the located store with hard binding.

A practical workflow for checking C2PA claim signature location, credential trust, algorithm support, and the final signature status.

A practical C2PA workflow for validating a trusted time-stamp while keeping signature time separate from capture time and event claims.

Learn how C2PA BMFF hash exclusions define covered MP4 bytes, how validators match paths and ranges, and what a mismatch can establish.

Review C2PA box order, identifiers, excluded boxes, and hash results for JPEG, PNG, and GIF without turning integrity into a truth verdict.

A file-based workflow for reading a C2PA data hash assertion, checking excluded byte ranges, and keeping asset integrity separate from claims about a depicted event.

A file-based workflow for identifying a compressed C2PA wrapper, recording decompression errors, and keeping container failure separate from manifest validity and real-world claims.

A file-based workflow for reading an update manifest separately from the original signed claim, checking allowed assertions, and keeping provenance records apart from real-world truth claims.

A file-based workflow for distinguishing a recorded C2PA redaction from an unexplained gap, while keeping provenance, content integrity, and real-world claims separate.

A practical workflow for recording the tool that created a C2PA claim without treating its name, version, or icon as proof that every assertion is correct.

A practical workflow for recording certificate trust, private credentials, time-stamp trust, and claim contents as separate checks.

Use a watermark or fingerprint to locate a detached C2PA Manifest, then validate the recovered record without treating the lookup as proof of integrity or truth.

A practical way to record C2PA success, informational, and failure codes without turning one technical result into a verdict about the media.

A repeatable workflow for copying recorded edit actions, preserving unknowns, and keeping signed provenance separate from claims about a real event.

Read ingredient records as a map of source assets, preserve the exact relationship and validation details, then verify the depicted event with separate evidence.

Learn what a Content Credential integrity check confirms, what a failure cannot prove, and how to preserve the exact file for further verification.

A valid Content Credential can establish an intact provenance record. The event shown still needs independent verification.

A repeatable workflow for checking a C2PA manifest, reading its assertions, recording unknowns, and keeping provenance separate from truth.

Pause a virtual-kidnapping ransom demand, compare the alleged proof-of-life media with confirmed photos, contact the person independently, and preserve the original evidence.

Use a separate contact path to verify an unexpected voice message before opening links, changing platforms, sharing information, or sending money.

Learn how to inspect a Content Credential, separate verified provenance from truth claims, and combine C2PA records with source checks and probabilistic detection.

A benchmark measures a system within a stated evaluation. Read its task and evidence before applying the result to an individual image or video.

A face and voice on a video call do not verify a participant or authorize a request. Use a separate channel to check the person, the action, and the approval before your team acts.

A practical workflow for finance and HR teams to verify urgent calls, meeting requests, payment changes, and identity claims before acting.

A cloned voice on the phone can sound right and still be synthetic. Using the FBI's fraud warning, here are practical checks for a suspicious payment or account request — and where a detector helps, and where it doesn't.

NIST's Open Media Forensics Challenge measures how detection algorithms find and locate image and video manipulation, but it doesn't verify one specific file. Here's how to read a detection score without overtrusting it.

Celebrity investment, giveaway and product-endorsement videos can use AI-cloned faces and voices. Using the FBI's fraud warning, here's a practical, source-first check for suspicious endorsements — and where a detector helps.

From August 2, 2026 the EU AI Act requires anyone deploying a deepfake to disclose it, with fines up to €15M or 3% of turnover. One marking duty was partially delayed to December — and none of it stops scammers. Here's what actually changes, in plain language.

Scammers can use AI-generated audio to imitate a relative and demand money. A voice that sounds right is no longer proof — use these steps to verify an urgent call before you act, and to check a recording.

News clips arrive stripped of context. This workflow traces where a file came from before running any detector, and treats detection scores as probability signals rather than verdicts.

A practical checklist for spotting manipulated or AI-generated video before sharing, including how probabilistic detection tools fit into your verification workflow.

A deepfake detection expert says he no longer trusts his own eyes — a warning that human judgment alone can't keep up with AI-generated media.

Detection scores are probabilistic, not verdicts. Learn how to use them alongside provenance and human review to make sounder judgments about digital media.

Savi raised $7M to launch an app detecting AI deepfake scams, including fake kidnapping ransom calls, on iPhone and Android.

Scam.ai partners with Qualcomm to launch Halo, an on-device deepfake detection model for video calls, unveiled at Computex 2026.

Google Gemini now offers free personalized AI image generation to US users, raising new deepfake and privacy concerns you must understand.

Google Cloud generative AI is transforming council planning operations by automating unstructured data across municipal agencies in 2026.

AI voice cloning scams trick victims into sending money by impersonating loved ones. Learn how these scams work and how to stay safe.

Deepfake detectors are powerful but not perfect. Learn what accuracy scores really mean and why no tool catches every fake.

AI fake audio calls are targeting banks and families in 2026 — learn how to identify AI-generated audio before it costs you.

Deepfake political videos are reshaping elections worldwide. Learn how they threaten democracy and how to spot them before they mislead you.

Learn how to detect AI-generated images with practical tips and free tools. Protect yourself from deepfakes in your social feed today.

Compare the best free deepfake audio detectors in 2026 to spot AI-cloned voices before they fool you or someone you love.

Documented deepfake fraud cases and official alerts with dates, amounts, media used, source links, and a practical verification workflow.

Learn 7 visual and audio signs that reveal AI-generated video, from glitchy backgrounds to unnatural voice timing.

AI-generated faces and voices are fueling a new wave of deepfake romance scams. Learn the warning signs and how to protect yourself today.

A practical C2PA hub for checking manifests, signatures, trust lists, actions, ingredients, and the boundary between provenance and real-world truth.

Compare file detectors, provenance verifiers, browser verification tools and detection APIs by the question they answer, the evidence they show and their limits.

Check a suspicious video with a source-backed workflow covering the original file, timeline, reverse search, Content Credentials and detector limits.

Use a private, source-first checklist to check suspicious Sabrina Carpenter images and videos without assuming a search result proves the file is fake.

Learn to identify sophisticated AI voice scams and deepfake clones used in phone fraud. This guide provides essential tips and detection methods to protect you from falling victim to synthetic voice deception.

With AI-generated images becoming increasingly sophisticated, it's crucial to distinguish between genuine photos and synthetic creations. Learn practical tips and tricks to investigate profile pictures and viral content for tell-tale signs of AI generation.

Learn to identify the common tell-tale signs of AI-generated images, from distorted hands to inconsistent lighting.