We Used to Need Video as Proof. Now the Video Needs Proof.

For most of the modern video era, the recording was the thing you brought out when words stopped being enough. A witness could misremember. Two people could tell different versions of the same incident. An official report could be challenged. Then someone would say there was video, and the argument changed. The camera did not settle every question, but it gave everyone something outside the competing stories to examine.

Artificial intelligence has complicated that arrangement in a particularly annoying way. The obvious problem is that fake video can now look increasingly plausible. The more consequential problem may be what happens to real video once everyone knows the fake version is possible. A recording can be authentic and still arrive under immediate suspicion. 'That is AI' becomes available not only as a warning about synthetic media, but as a way to dismiss inconvenient evidence before anyone has established whether it was altered at all.

SWEAR, a Boise-based digital-content-authenticity company, is responding to that second problem with something refreshingly practical. On Sept. 10, it announced the Community Video Integrity Project, a program intended to help cities, law-enforcement agencies and other public-sector organizations establish a cryptographically verifiable record of selected video from the moment it is captured. The basic idea is not to ask an algorithm later whether a clip looks fake. It is to create proof at the beginning that can be checked later if the recording is challenged.

That distinction matters. Deepfake detection is necessarily a game of inspection: analyze the file after the fact and estimate whether something about it looks synthetic or manipulated. SWEAR is trying to move the trust question upstream. According to the company, its system creates cryptographic fingerprints for video frames and audio during capture, anchors the resulting proof data to a permissioned, tamper-evident ledger, and later compares a recording against that original proof. If the file has changed, the validation fails or identifies the mismatch. The company says the media itself is not stored on the ledger; the stored material is the proof data used to verify it.

The Community Video Integrity Project is aimed at public agencies already living with a very specific operational reality. Municipal cameras, security systems and investigative video can become part of an incident review, a criminal case, a civil dispute, a public-records release or an argument about government accountability. SWEAR says selected participants will deploy its technology on high-priority cameras, including in environments using Milestone Systems' XProtect video-management platform. In practical terms, the project treats authenticity as another part of evidence management rather than something an agency scrambles to establish only after somebody cries fake.

That feels like the right direction because the burden is changing. A 2025 Pew Research Center survey found that 53 percent of U.S. adults were not too confident or not at all confident that they could tell whether pictures, video and text were made by AI or by people. Seventy-six percent said being able to tell the difference was extremely or very important. Those two numbers belong together. People care about the distinction, but many do not trust themselves to make it by looking. The answer cannot simply be 'look harder.'

This is where the deeper structural problem appears. We used to think of authenticity as something visible in the object itself. A photograph looked like a photograph. A recording sounded like a recording. A surveillance clip came out of a surveillance system and was presumed to have a relationship to an actual camera pointed at an actual place. Generative systems break that intuitive shortcut. Once fabrication becomes ordinary enough, visual plausibility stops doing the authentication work. The file needs a history.

The media world is already moving in that direction. The Coalition for Content Provenance and Authenticity, or C2PA, has developed technical standards for attaching verifiable provenance information to digital media so users can inspect where a piece of content came from and what happened to it. SWEAR takes a different technical approach and says it does not directly follow C2PA, but the larger impulse is the same: trust can no longer depend only on what an image or recording appears to be. We are beginning to build infrastructure around the question of where it came from, how it was handled and whether it changed along the way.

There is an important limitation hidden inside the word 'authentic,' and understanding it actually makes these systems more useful. Cryptographic verification can establish that a particular recording matches the material captured by a particular system and has not been altered afterward. It cannot tell you what an event means. It cannot determine whether the camera angle was misleading, whether something happened outside the frame, whether a scene was staged before recording began, or whether an official interpretation of the footage is correct. Provenance is not omniscience. What it can do is preserve one increasingly valuable fact: this is the recording we actually captured, and this is what happened to the file afterward.

That narrower promise may become essential because synthetic media creates what legal scholars Bobby Chesney and Danielle Citron famously called the 'liar's dividend.' The concept is simple: once people understand that convincing fake audio and video can exist, dishonest actors gain a new defense against authentic material. They do not necessarily need to manufacture a fake. They can benefit from the existence of fakes by claiming the real thing is synthetic. The Brennan Center has documented versions of this dynamic in politics and litigation, including lawyers attempting what has been described as a deepfake defense against audiovisual evidence.

The phrase sounds academic until you translate it into ordinary civic life. Imagine a contested police encounter, a dispute over what happened inside a municipal building, footage from a transit station, a traffic incident, a public demonstration or a security-camera recording that becomes important after an emergency. In the old argument, the question might have been whether the video showed what one side said it showed. The new argument can begin one level earlier: is that even the real video? A person trying to delay, confuse or discredit the evidence does not have to prove manipulation. Sometimes planting the possibility is useful enough.

That changes the economics of doubt. Saying 'that could be AI' is cheap. Proving that a recording is authentic after the challenge has been raised can be expensive, slow and technically complicated. Agencies may need forensic review, device records, export logs, chain-of-custody documentation and expert testimony simply to get back to the place where everyone once started: yes, this is the actual footage. Capture-time verification tries to reverse that asymmetry by creating the authenticity record before anyone knows which recording will later matter.

That is why SWEAR's project is more interesting than another corporate announcement about fighting deepfakes. It treats the problem as infrastructure. Nobody knows which camera will capture the incident that becomes important next month. Nobody knows which routine recording will eventually be challenged in court or debated in public. If authenticity protection is only applied after a clip becomes controversial, the strongest moment to establish provenance has already passed. The practical answer is to make the proof boring, automatic and present from the start.

There is something deeply strange about needing another technology layer to certify reality because technology made reality easier to counterfeit. We built tools capable of manufacturing convincing images, voices and video, and one of the consequences is that we now need cryptography, secure ledgers and provenance systems to restore some of the trust that cameras once received by default. It is a technological arms race, but not only between fake generation and fake detection. It is also a race to preserve the credibility of the real thing.

For public agencies, that credibility matters beyond courtrooms. Video increasingly operates in the space between institutions and the people they serve. A city releases footage after an incident because the footage is supposed to show its work. A police department presents body-worn or surveillance video because the recording is expected to provide a factual reference point. Investigators rely on cameras because the record can outlast memory. If every important clip can be neutralized by a casual allegation that it was generated or altered, the damage is not confined to one case. The shared evidentiary floor gets weaker.

The goal, then, should not be to return to a naive era in which people believed everything a camera showed. Cameras have always had angles, omissions, operators and contexts. Skepticism is healthy. But skepticism works best when it has something to examine. Provenance gives it structure. Instead of arguing from vibes about whether a video 'looks real,' people can ask a more useful set of questions: Which device captured it? When was the proof created? Has the file changed? Does the presented copy match the protected record? Those are answerable questions.

SWEAR's Community Video Integrity Project will not solve synthetic media, and no authenticity system can force people to accept evidence they are determined to reject. What it represents is a practical adaptation to the world we have already built. When fabrication becomes cheap, verification has to become routine. When denial becomes easy, institutions need stronger records.

And when 'here is the video' no longer ends the authenticity argument, the next sentence increasingly needs to be: here is the proof that this is the video.

SOURCE NOTES

• SWEAR, “SWEAR Launches Program to Help Cities Protect the Integrity of Critical Video Evidence,” Sept. 10, 2026
• SWEAR, “Technology You Can Trust When Truth Is on the Line”
• SWEAR Support, “How does SWEAR work?”
• Pew Research Center, “How Americans View AI and Its Impact on Human Abilities, Society,” Sept. 17, 2025
• Brennan Center for Justice, “Deepfakes, Elections, and Shrinking the Liar’s Dividend,” Jan. 23, 2024
• Coalition for Content Provenance and Authenticity, C2PA Specifications and Guiding Principles