The Danish Gambit, Part 5
How Transparency Would Benefit All but the Bad Actors
When GPT and Claude flagged deepfakes’ utility for blackmail and extortion, I realized teenagers are precisely the demographic Denmark’s production exemption most endangers. Even without deepfakes or photographic material, teens already face online extortion. When your social circle is your whole world and you lack adults’ confidence that people will know better, the leverage potential is devastating. A friend creates a compromising deepfake, stores it privately, and suddenly has permanent ammunition. The victim can’t prove it’s fake because the technology is indistinguishable from reality, and the social damage is immediate regardless of truth.
The artistic expression defense for deepfakes collapses under scrutiny. We don’t teach students to faithfully reproduce dollar bills—that would normalize counterfeiting as legitimate skill-building. As Gemini pointed out, even art museums prohibit 1:1 reproductions of masterworks. If replicating paintings and currency is off-limits despite requiring genuine artistic skill, why should realistic identity replication be protected just because it’s digital? The technology to create deepfakes may be impressive, but so is the craft of counterfeiting. We’ve already decided that certain capabilities are too dangerous to permit regardless of the skill involved.
Claude’s suggestion that teens could “safely” experiment with deepfakes of themselves gave me pause. Teens don’t practice security hygiene, making them prime targets for device theft and hacking. A self-deepfake created for harmless experimentation becomes blackmail material the moment their phone is stolen, a friend becomes an ex-friend with access to old files, or family members discover it and react with judgment. The teen “consented” to creating it but couldn’t possibly consent to all the ways it might escape their control. Even self-directed deepfakes represent ongoing vulnerability that teens lack the foresight to manage.
There’s a darker consequence I hadn’t initially considered: reality calibration. I remember drawing a tiger as a kid and discovering my hands wouldn’t produce the image in my head. That gap between imagination and execution taught me reality has constraints—physics, biology, skill limitations. Today’s kids, saturated with video games and deepfakes showing impossible physics, are being robbed of those reality-check moments. When teenagers see deepfakes of peers surviving falls that would kill them or performing stunts no human body can execute, they’re learning that physical limits are negotiable. We’ve already seen TikTok challenge deaths from kids attempting dangerous stunts. Now imagine those challenges “documented” with deepfakes showing successful completion when the real attempt would be fatal.
The teen vulnerability cases all share a common thread with a seemingly unrelated problem: scientific image contamination. AI-generated images of non-existent species are flooding databases researchers rely on for identification and classification. The connection isn’t obvious until you see it: in both cases, synthetic content masquerading as authentic creates cascading harm. Victims of impersonation lose control of their identity. Research gets contaminated by fake data that researchers can’t distinguish from legitimate specimens. Kids develop distorted understandings of what’s physically possible. The solution across all these domains isn’t banning the technology—it’s requiring honest disclosure about what’s synthetic and what’s real.
Mandatory labeling solves multiple problems simultaneously. Scientists searching for species images can filter out AI-generated content and work with authentic specimens. Platforms can automatically verify and mark synthetic material, giving users clear signals about what they’re viewing. Kids learning to navigate media can see permanent markers distinguishing real from synthetic, helping them develop accurate reality models. And victims of deepfakes have legal recourse not just when content is published, but when unlabeled synthetic versions of them are created without consent. Transparency doesn’t prevent all harm, but it prevents the specific harm that comes from deception—the harm deepfakes are designed to create.
Denmark’s framework is sophisticated and carefully considered, but it needs transparency requirements matching its production standards. If you create synthetic content depicting identifiable people, it must carry permanent, visible, tamper-resistant labels embedded in both metadata and visual presentation. These labels should survive re-encoding, cropping, and format conversion—they’re not decorative watermarks users can strip away but cryptographic signatures baked into the file structure itself. This is technically feasible; the C2PA standard already exists and camera manufacturers are implementing it. The infrastructure is ready; what’s missing is legal mandate.
This isn’t censorship or restriction of legitimate creative work. It’s honest disclosure, the same principle we apply to food ingredients, securities filings, and pharmaceutical labeling. We don’t ban complex financial instruments; we require disclosure so investors can make informed decisions. We don’t ban chemicals in food, requiring instead ingredient lists so consumers can protect themselves. Calling for authenticity labels on media follows the same logic: give people the information they need to assess what they’re seeing. Artists can still express themselves, creators can still generate synthetic images for legitimate purposes, and platforms can still host user content. They just have to be honest about what’s real and what’s not.
The only people harmed by transparency requirements are those whose business model or agenda depends on deception. Legitimate creators have nothing to hide: they want audiences to know their work is creative interpretation, not fraudulent impersonation. Researchers generating synthetic species images for modeling purposes benefit from clear labeling that prevents their work from contaminating authentic databases. But extortionists, fraudsters, and those seeking to manipulate public perception through fake evidence lose their most effective tool: the assumption that seeing is believing. That’s exactly who should be disadvantaged by this policy. Denmark got the copyright framework right. Now it needs to close the production loophole and add transparency requirements. Together, those changes would create the comprehensive protection people actually need.
[This post was drafted with assistance from Claude Sonnet 4.5, following conversations with ChatGPT-5.1 Instant, Gemini 3 Flash/Pro, and Claude Sonnet 4.5.]
GPT-5.1 Instant
Prompt: Right. And we hear of cases where blackmailers tried to extort teens after getting them to overshare and then threatening they’d show that material to their parents or friends. Deepfakes would be just as terrifying for the prospective victims, especially younger ones who lack the confidence that their friends would know that they’d never do such a thing and won’t fall for the fabricated version.
Claude Sonnet 4.5
Prompt: Yes, they could now get more realistic versions of Jackass where the stunts actually work out as they pictured in their minds.
You combine that with video game culture (which already weakens gamers’ calibration of reality and practical goal setting) and parents are going to have a really hard time keeping their kids grounded in reality. It was a while ago, but a journalist who has covered MBS for years speculated that his unrealistic approach to that city he was building might be attributed to how much time he was spending playing video games (and making unrealistic demands of the tech teams because he thought reality should work out just as easily as the video game version does).
Although it’s been decades since I was a four-year-old, one memory I still have from those years is drawing a tiger and realizing how even my hands, which are probably the parts I can exercise the most control over, cannot be manipulated to produce exactly what I envisioned. With kids today being much less exposed to such experiences where they can reality-check themselves, we might need to keep those hands-on classroom activities for younger children so they don’t confuse what they see on their screens with reality.
Gemini 3 Pro
Prompt: Do you think we’ve given deepfakes a fair shake (even though I called that my “legal weasel” moment)? The more I think about them, I can only surface bad-faith scenarios. All the problems arise because of the lack of transparency, which we need to get much more proactive on. Transparent disclosures would also help with photorealistic AI-synthesized specimens that scientists complain about finding in their image searches.
