A fake AI-generated image showing a US airman surrounded by smiling military members after an alleged rescue from Iran has been shared more than 21,000 times on X, including by Republican officials. The incident is a clear demonstration that we're past the point where AI fakes are a novelty—they're now a real information crisis. According to The Guardian, the image purported to show a crew member who had been rescued from Iran, surrounded by celebrating military personnel. It looked plausible enough that elected officials shared it without verification, and it spread rapidly through conservative social media circles. Except it never happened. The image was generated by AI. We've known for years that AI could create fake images. What's changed is the quality and the scale. Early AI-generated images had telltale signs—weird fingers, distorted faces, impossible lighting. Modern AI image generators have gotten better, and the errors are more subtle. More importantly, the images spread faster than fact-checks can catch them. When an elected official shares a fake image to tens of thousands of followers, and those followers share it to their networks, you can have tens of thousands of shares before anyone notices it's fake. By then, the damage is done. Even when corrections are posted, many people who saw the original never see the correction. This isn't just about partisan politics, though the fact that it spread among Republican officials is notable given the current political environment around Iran. The deeper problem is epistemic: how do we know what's real anymore? For most of human history, photographs were evidence. "Pics or it didn't happen" was a reasonable standard. If you saw a photo, you could generally trust that someone was there with a camera and captured that moment. AI has destroyed that assumption. Now a photo might be real, or it might be generated in seconds by someone with access to an image generator and an agenda. The technology to detect AI-generated images exists, but it's not foolproof, and it's not built into the platforms where these images spread. Journalists and fact-checkers are overwhelmed. For every fake image that gets debunked, dozens more circulate without scrutiny. Social media platforms could implement detection systems, require disclosures for AI-generated content, or slow the spread of unverified images. But that would conflict with their engagement metrics. Viral content—real or fake—drives usage, and usage drives ad revenue. The incentives are misaligned. We're essentially in an information environment where seeing is no longer believing, but most people haven't caught up to that reality. They still treat images as evidence, especially when they confirm what they already believe or want to be true. That's a dangerous combination. Politicians sharing fake images is embarrassing, but it's also a preview of what's to come. If elected officials can't tell real from fake, how can we expect ordinary citizens to do better? And if we can't trust images, what's next? AI-generated video is already here, and it's only getting better. The question isn't whether AI can fool people. It clearly can. The question is what we're going to do about it—and whether we'll do it before the problem gets much worse.