Courts weigh social media evidence rules

Social media posts used to look easy in court — grab the screenshot, argue about what it means, move on. That is not the world judges are in now. A post might be real but ambiguous. A screenshot might be fake. An AI output might look scientific while hiding a black box underneath. And a criminal case can turn on exactly those gaps. That is why this story is bigger than one weird image or one prosecutor’s theory. The pressure is coming from both directions at once — social posts are being treated as evidence of intent, while AI is making digital evidence much easier to forge or overread. ### Why is a social post suddenly the test case? Because the clearest current example is the James Comey case. On April 28, 2026, federal prosecutors indicted the former FBI director over an Instagram photo showing seashells arranged as “8647,” arguing the image was a threat against President Donald Trump. Comey denies wrongdoing, and the case turns a lot on intent — did the post communicate a real threat, or did prosecutors layer threatening meaning onto an ambiguous image after the fact? (cnbc.com) ### Why is intent the hard part? A threat case is not just about whether words or symbols can be read violently. Prosecutors generally need to show the defendant knowingly and willfully communicated a threat. That gets messy fast with online culture, where numbers, memes, irony, and in-group slang can mean different things to different audiences. In the Comey prosecution, legal analysts are already saying that proving that mental state could be the hardest part. (cnbc.com) ### Isn’t a screenshot enough? Not anymore — basically because “digital” no longer means “self-authenticating.” Courts have always required relevance and authentication, but AI raises the cost of both. A screenshot can be edited. A chat log can be fabricated. A video can show a real-looking person saying things they never said. That means the old move — print the post and let the jury sort it out — is getting less defensible. (thehill.com) ### What changed with AI evidence? Judges started seeing actual deepfakes and fabricated digital exhibits show up in live cases. One of the most cited examples is Mendones v. Cushman & Wakefield in California, where a judge identified AI-generated witness video submitted as if it were authentic testimony. The case was dismissed on Sept. 9, 2025, and the judge later denied reconsideration on Nov. 6. That was a warning shot — not a theory, an actual courtroom problem. (ncsc.org) ### So what are courts doing about it? The federal rules process is trying to build a gate before juries ever see some of this material. Proposed Rule 707 would make certain machine-generated evidence clear the same reliability hurdles that expert testimony faces under Rule 702. In plain English, if software spits out an analysis that acts like expert proof, the party offering it may have to show the data, method, and application are reliable rather than just saying “the computer did it.” (ncsc.org) ### Where do AI prompts fit? They create a second problem. Sometimes the AI output itself is the exhibit. Other times the prompt history matters because it may show planning, fabrication, or consciousness of guilt — like asking a model to generate a fake message, image, or forensic-looking explanation. But prompt logs raise their own fights over authorship, completeness, hearsay, and whether the machine’s answer is being offered for truth or just to show what the user was trying to do. (natlawreview.com) That is new terrain, and courts are still feeling for the edges. ### Why does this matter beyond one case? Because threat prosecutions are already rising — one count cited by legal experts put federal threat cases against public officials at 133 in 2025. Add AI-made evidence to that environment and the risk is obvious: more cases built from symbols, screenshots, and machine outputs that look persuasive before anyone proves they are real. (natlawreview.com) ### Bottom line? The fight is shifting from “is this relevant?” to “what exactly is this?” For social posts, that means intent and context. For AI material, that means authenticity and reliability. And for judges, the job is getting much more technical before the jury ever walks in. (ncsc.org) (thehill.com)