{"slug": "synthid", "name": "SynthID", "category": "ai", "category_name": "AI models and assistants", "description": "", "recommenders": 4, "commenters": 26, "warners": 10, "score": -2.712, "rising": 1.35, "first": "2025-11-20", "latest": "2026-10-07", "why": "SynthID is a better standard for designating something was created with GenAI", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 4, "neutral": 9, "compare": 3, "warn": 12}, "spark": [0, 0, 2, 0, 0, 0, 0, 0, 0, 0, 1, 1], "listed": true, "loved": [], "caveat": "", "short_quote": "SynthID is a better standard for designating something was created with GenAI", "kind": "", "icon": true, "domain": "deepmind.google", "github_repo": null, "official_url": "https://deepmind.google/models/synthid/", "pricing": "", "aliases": ["SynthID"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 799278544, "item_id": 49954909, "url": "https://news.ycombinator.com/item?id=49954909", "by": "dagmx", "date": "2026-10-04", "stance": "recommend", "quote": "SynthID is a better standard for designating something was created with GenAI", "excerpt": "The EU AI act doesn\u2019t force C2PA or any particular standard. Yes many people will use C2PA erroneously, but C2PA is not meant for indicating content was virtually generated, but that it was not modified from origination. C2PA is fallible if your origin is fallible. SynthID is a better standard for designating something was created with GenAI, though is limited in what it covers. For example, 3D content is currently a", "context": "", "use_case": "", "switched_from": "", "story": "How to hack time, with C2PA"}, {"id": 796717920, "item_id": 49794870, "url": "https://news.ycombinator.com/item?id=49794870", "by": "pavo-etc", "date": "2026-09-21", "stance": "recommend", "quote": "Tech like SynthID I see a net positive", "excerpt": "I'm not convinced spymark is better than just \"invisible watermarks\", spymark to my ears sounds designed to be sound very negative when invisible watermarks are not always negative, e.g. the counterfeit bank note example. Tech like SynthID I see a net positive especially since it doesn't degrade text quality. I dream about a browser extension running at all times that makes text more translucent based on the confiden", "context": "", "use_case": "", "switched_from": "", "story": "Spymarks, not Watermarks"}, {"id": 744356960, "item_id": 46522310, "url": "https://news.ycombinator.com/item?id=46522310", "by": "aprilnya", "date": "2026-01-06", "stance": "recommend", "quote": "It is, because of SynthID.", "excerpt": "It is, because of SynthID. Checked it myself: https://gemini.google.com/share/825ee3c53dd8 More info about SynthID: https://deepmind.google/models/synthid/", "context": "", "use_case": "", "switched_from": "", "story": "Debunking the AI food delivery hoax that fooled Reddit"}, {"id": 743752465, "item_id": 46484529, "url": "https://news.ycombinator.com/item?id=46484529", "by": "satvikpendem", "date": "2026-01-03", "stance": "recommend", "quote": "specific watermarking in the images themselves", "excerpt": "> And for what it's worth, the image pretty consistently trips gen AI image detectors (e.g., https://hivemoderation.com/ai-generated-content-detection). Notwithstanding your previous points, these sorts of AI detectors flag many false positives, they're not worth relying on. The only one that could actually work is specific watermarking in the images themselves, such as what Google does with SynthID in their generate", "context": "", "use_case": "", "switched_from": "", "story": "Slop Is Slop"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 799902384, "item_id": 49993899, "url": "https://news.ycombinator.com/item?id=49993899", "by": "m00dy", "date": "2026-10-07", "stance": "warn", "quote": "Deepwalker cracked this one long time ago", "excerpt": "Deepwalker cracked this one long time ago https://deepwalker.xyz/blog/evaluating-synthid-watermark-robustness", "context": "", "use_case": "", "switched_from": "", "story": "SynthID Detector"}, {"id": 799902096, "item_id": 49993881, "url": "https://news.ycombinator.com/item?id=49993881", "by": "Retr0id", "date": "2026-10-07", "stance": "warn", "quote": "It's a real shame Google isn't more transparent about how it actually works", "excerpt": "It's a real shame Google isn't more transparent about how it actually works, and doesn't provide any mechanism for classifying images in bulk or offline. This is the best SynthID write-up I've found so far: https://fyx.me/articles/attempting-model-extraction-of-google-deepmind-synthid-image-watermark-detector-and-exploring-adversarial-machine-learning-attacks-against-synthid/ It covers how it actually works (probably", "context": "", "use_case": "", "switched_from": "", "story": "SynthID Detector"}, {"id": 796809696, "item_id": 49800606, "url": "https://news.ycombinator.com/item?id=49800606", "by": "mpalmer", "date": "2026-09-22", "stance": "warn", "quote": "It absolutely does. It constrains high-entropy word choice so it can \"store\" other things in your text.", "excerpt": "I'm not convinced spymark is better than just \"invisible watermarks\", spymark to my ears sounds designed to be sound very negative when invisible watermarks are not always negative, e.g. the counterfeit bank note example. The article spent quite a bit of energy explaining why the word choice, seems like you're just ignoring that? Also, bank note watermarks are not invisible. Watermarks are not invisible, as the artic", "context": "", "use_case": "text watermarking", "switched_from": "", "story": "Spymarks, not Watermarks"}, {"id": 786064336, "item_id": 49129021, "url": "https://news.ycombinator.com/item?id=49129021", "by": "ncr100", "date": "2026-07-31", "stance": "warn", "quote": "Of course that's impractical because a million people could quickly receive the image", "excerpt": "Does this move us closer to needing registries of authentic images? (White-listing? Only a partial solution.) Google's response today, in addition to removing nano banana from Earth, was that people should just use Google's synthID(?) AI detection app, to determine if this is a blacklisted (effectively) image. Of course that's impractical because a million people could quickly receive the image .. the image would mak", "context": "", "use_case": "", "switched_from": "", "story": "Nano Banana 2 removed from Google Earth"}]}], "who": [], "tags": [], "use_cases": ["text watermarking", "llm text watermarking", "ai image watermark bypass"], "compared": [{"slug": "gemini", "name": "Gemini", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=48198291", "title": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool", "mentions": 6}, {"url": "https://news.ycombinator.com/item?id=49993188", "title": "SynthID Detector", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49794615", "title": "Spymarks, not Watermarks", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49324087", "title": "Anthropic's 'Watermark' Text Adulteration in Claude Is a Perversion of Writing", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=47709130", "title": "Discovering, detecting, and surgically removing Google's AI watermark", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=47169146", "title": "SynthID", "mentions": 2}], "sentiment_by_month": {"2025-11": {"neutral": 1, "compare": 1}, "2025-12": {}, "2026-01": {"recommend": 2, "neutral": 1}, "2026-02": {"neutral": 3, "warn": 1}, "2026-03": {}, "2026-04": {"compare": 1, "warn": 2}, "2026-05": {"neutral": 1, "warn": 5, "compare": 1}, "2026-06": {}, "2026-07": {"warn": 1}, "2026-08": {"neutral": 3}, "2026-09": {"recommend": 1, "warn": 1}, "2026-10": {"recommend": 1, "warn": 2}}, "quotes": [{"id": 799902384, "item_id": 49993899, "url": "https://news.ycombinator.com/item?id=49993899", "by": "m00dy", "date": "2026-10-07", "stance": "warn", "quote": "Deepwalker cracked this one long time ago", "excerpt": "Deepwalker cracked this one long time ago https://deepwalker.xyz/blog/evaluating-synthid-watermark-robustness", "context": "", "use_case": "", "switched_from": "", "story": "SynthID Detector"}, {"id": 799902096, "item_id": 49993881, "url": "https://news.ycombinator.com/item?id=49993881", "by": "Retr0id", "date": "2026-10-07", "stance": "warn", "quote": "It's a real shame Google isn't more transparent about how it actually works", "excerpt": "It's a real shame Google isn't more transparent about how it actually works, and doesn't provide any mechanism for classifying images in bulk or offline. This is the best SynthID write-up I've found so far: https://fyx.me/articles/attempting-model-extraction-of-google-deepmind-synthid-image-watermark-detector-and-exploring-adversarial-machine-learning-attacks-against-synthid/ It covers how it actually works (probably", "context": "", "use_case": "", "switched_from": "", "story": "SynthID Detector"}, {"id": 799278544, "item_id": 49954909, "url": "https://news.ycombinator.com/item?id=49954909", "by": "dagmx", "date": "2026-10-04", "stance": "recommend", "quote": "SynthID is a better standard for designating something was created with GenAI", "excerpt": "The EU AI act doesn\u2019t force C2PA or any particular standard. Yes many people will use C2PA erroneously, but C2PA is not meant for indicating content was virtually generated, but that it was not modified from origination. C2PA is fallible if your origin is fallible. SynthID is a better standard for designating something was created with GenAI, though is limited in what it covers. For example, 3D content is currently a", "context": "", "use_case": "", "switched_from": "", "story": "How to hack time, with C2PA"}, {"id": 796809696, "item_id": 49800606, "url": "https://news.ycombinator.com/item?id=49800606", "by": "mpalmer", "date": "2026-09-22", "stance": "warn", "quote": "It absolutely does. It constrains high-entropy word choice so it can \"store\" other things in your text.", "excerpt": "I'm not convinced spymark is better than just \"invisible watermarks\", spymark to my ears sounds designed to be sound very negative when invisible watermarks are not always negative, e.g. the counterfeit bank note example. The article spent quite a bit of energy explaining why the word choice, seems like you're just ignoring that? Also, bank note watermarks are not invisible. Watermarks are not invisible, as the artic", "context": "", "use_case": "text watermarking", "switched_from": "", "story": "Spymarks, not Watermarks"}, {"id": 796717920, "item_id": 49794870, "url": "https://news.ycombinator.com/item?id=49794870", "by": "pavo-etc", "date": "2026-09-21", "stance": "recommend", "quote": "Tech like SynthID I see a net positive", "excerpt": "I'm not convinced spymark is better than just \"invisible watermarks\", spymark to my ears sounds designed to be sound very negative when invisible watermarks are not always negative, e.g. the counterfeit bank note example. Tech like SynthID I see a net positive especially since it doesn't degrade text quality. I dream about a browser extension running at all times that makes text more translucent based on the confiden", "context": "", "use_case": "", "switched_from": "", "story": "Spymarks, not Watermarks"}, {"id": 789351888, "item_id": 49334493, "url": "https://news.ycombinator.com/item?id=49334493", "by": "mangoman", "date": "2026-08-17", "stance": "neutral", "quote": "actually reading the the synthid paper showed me that my understanding was wrong", "excerpt": "huh, thanks for commenting this! I trusted the linked explanation post https://declaude.org/watermarking/ but actually reading the the synthid paper showed me that my understanding was wrong: https://www.nature.com/articles/s41586-024-08025-4 It is definitely blurrier whether you can say this approach changes the distribution then. By definition, it _has_ to change the probabilities of output tokens, but it's not tot", "context": "", "use_case": "", "switched_from": "", "story": "Anthropic's 'Watermark' Text Adulteration in Claude Is a Perversion of Writing"}, {"id": 789296704, "item_id": 49331044, "url": "https://news.ycombinator.com/item?id=49331044", "by": "bccdee", "date": "2026-08-17", "stance": "neutral", "quote": "SynthID should not distort the output of an LLM any more than the default temperature settings do already.", "excerpt": "> \u201cHe leaped at the chance\u201d and \u201cHe jumped at the opportunity\u201d are very similar sentences expressing the same general sentiment, but they are not the same. The exact words we choose when writing matter. I want any LLM I use to choose the very best, most precise words at every single decision point. Neither of these is \"better\" or \"more precise\"; in fact, LLMs will generally choose randomly between these candidates ba", "context": "", "use_case": "LLM text watermarking", "switched_from": "", "story": "Anthropic's 'Watermark' Text Adulteration in Claude Is a Perversion of Writing"}, {"id": 788726576, "item_id": 49295411, "url": "https://news.ycombinator.com/item?id=49295411", "by": "giridharkannan", "date": "2026-08-13", "stance": "neutral", "quote": "it feels like it will be difficult to remove watermarking on bigger text", "excerpt": "https://declaude.org/watermarking/ did a good job in explaining how SynthID works. As per their blog, it feels like it will be difficult to remove watermarking on bigger text and the checking for watermarking is also not complex", "context": "", "use_case": "", "switched_from": "", "story": "Text AI watermarks will always be trivial to remove"}, {"id": 786064336, "item_id": 49129021, "url": "https://news.ycombinator.com/item?id=49129021", "by": "ncr100", "date": "2026-07-31", "stance": "warn", "quote": "Of course that's impractical because a million people could quickly receive the image", "excerpt": "Does this move us closer to needing registries of authentic images? (White-listing? Only a partial solution.) Google's response today, in addition to removing nano banana from Earth, was that people should just use Google's synthID(?) AI detection app, to determine if this is a blacklisted (effectively) image. Of course that's impractical because a million people could quickly receive the image .. the image would mak", "context": "", "use_case": "", "switched_from": "", "story": "Nano Banana 2 removed from Google Earth"}, {"id": 771246000, "item_id": 48202875, "url": "https://news.ycombinator.com/item?id=48202875", "by": "827a", "date": "2026-05-19", "stance": "warn", "quote": "SynthID has been totally busted open", "excerpt": "Interesting that it seems to be the case that SynthID has been totally busted open, but OpenAI's new watermark has not yet [1] [1] https://github.com/wiltodelta/remove-ai-watermarks", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 771242128, "item_id": 48202633, "url": "https://news.ycombinator.com/item?id=48202633", "by": "m00dy", "date": "2026-05-19", "stance": "warn", "quote": "It\u2019s definitely hackable", "excerpt": "It\u2019s definitely hackable, Some of our engineers worked on this long time ago https://deepwalker.xyz/blog/bypassing-synthid-in-gemini-photos", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 771239024, "item_id": 48202439, "url": "https://news.ycombinator.com/item?id=48202439", "by": "mpetrovich", "date": "2026-05-19", "stance": "warn", "quote": "Seems inferior to C2PA", "excerpt": "Seems inferior to C2PA, which is actually an open standard: https://contentauthenticity.org/", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 771207936, "item_id": 48200496, "url": "https://news.ycombinator.com/item?id=48200496", "by": "echelon", "date": "2026-05-19", "stance": "warn", "quote": "SynthID payloads work the same way", "excerpt": "I'm not going to mince words - what you're saying is dangerous and harmful. > to be clear can not track you in any way All they have to do is encode enough entropy for a database unique identifier. Systems like this have been used to do it for audio: https://github.com/swesterfeld/audiowmark SynthID payloads work the same way, and the paper discusses encoding a \"user identifier\": https://arxiv.org/html/2510.09263v1#S", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 771188656, "item_id": 48199291, "url": "https://news.ycombinator.com/item?id=48199291", "by": "programd", "date": "2026-05-19", "stance": "compare", "quote": "There is an approach that apparently works and is based on spectral analysis of the images.", "excerpt": "Define easily. There is an approach that apparently works and is based on spectral analysis of the images. https://github.com/aloshdenny/reverse-SynthID", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 771186608, "item_id": 48199163, "url": "https://news.ycombinator.com/item?id=48199163", "by": "flaxxer", "date": "2026-05-19", "stance": "warn", "quote": "https://synthidbypass.com/", "excerpt": "https://synthidbypass.com/", "context": "", "use_case": "AI image watermark bypass", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 771177648, "item_id": 48198603, "url": "https://news.ycombinator.com/item?id=48198603", "by": "svieira", "date": "2026-05-19", "stance": "neutral", "quote": "and the previous SynthID: https://deepmind.google/blog/identifying-ai-generated-images-with-synthid/", "excerpt": "Now that they've broken the ability to trust video, they're looking to build it back, as long as you're allowed to use the tools: https://blog.google/innovation-and-ai/products/identifying-ai-generated-media-online/ (and the previous SynthID: https://deepmind.google/blog/identifying-ai-generated-images-with-synthid/) But it very much is \"close the barn door after the horse has bolted and the barn has otherwise burned", "context": "", "use_case": "", "switched_from": "", "story": "Gemini Omni"}, {"id": 763403024, "item_id": 47712689, "url": "https://news.ycombinator.com/item?id=47712689", "by": "alwa", "date": "2026-04-09", "stance": "warn", "quote": "a meaningless signal like this deserves such a demise", "excerpt": "As gp suggests, it\u2019s meaningless. And the frustrating/dangerous part is that it\u2019s holding itself out as meaningful to laypeople (including regulators). \u201cIt has a watermark, so it must be Gemini\u2019s work?\u201d Not in a particularly strong sense: if you want a watermark on whatever image, you can do what the repo authors did and yeet it through Gemini with \u201creproduce this exactly.\u201d \u201cIt doesn\u2019t have a watermark, so it must be", "context": "", "use_case": "", "switched_from": "", "story": "Discovering, detecting, and surgically removing Google's AI watermark"}, {"id": 763391744, "item_id": 47711984, "url": "https://news.ycombinator.com/item?id=47711984", "by": "m00dy", "date": "2026-04-09", "stance": "warn", "quote": "It's not really hard as it looks", "excerpt": "It's not really hard as it looks https://deepwalker.xyz/blog/bypassing-synthid-in-gemini-photos", "context": "", "use_case": "", "switched_from": "", "story": "Discovering, detecting, and surgically removing Google's AI watermark"}, {"id": 763379616, "item_id": 47711226, "url": "https://news.ycombinator.com/item?id=47711226", "by": "therobots927", "date": "2026-04-09", "stance": "compare", "quote": "", "excerpt": "https://deepmind.google/models/synthid/", "context": "", "use_case": "", "switched_from": "", "story": "How Do You Find an Illegal Image Without Looking at It?"}, {"id": 754729376, "item_id": 47170586, "url": "https://news.ycombinator.com/item?id=47170586", "by": "elpocko", "date": "2026-02-26", "stance": "warn", "quote": "SynthID is removed from images by passing the image through a diffusion model", "excerpt": "It doesn't. I don't have a link for you right now but there was a post on reddit recently showing that SynthID is removed from images by passing the image through a diffusion model for a single step at low denoise. The output image is identical to the input image (to the human eye).", "context": "", "use_case": "", "switched_from": "", "story": "SynthID"}, {"id": 754719760, "item_id": 47169985, "url": "https://news.ycombinator.com/item?id=47169985", "by": "squigz", "date": "2026-02-26", "stance": "neutral", "quote": "Looks like there's a lot more info here, at least about the text version.", "excerpt": "Looks like there's a lot more info here, at least about the text version. https://ai.google.dev/responsible/docs/safeguards/synthid", "context": "", "use_case": "", "switched_from": "", "story": "SynthID"}, {"id": 753881968, "item_id": 47117623, "url": "https://news.ycombinator.com/item?id=47117623", "by": "daemonologist", "date": "2026-02-22", "stance": "neutral", "quote": "Google watermarks their images", "excerpt": "Google watermarks their images; I believe it's a sort of subtle rainbow striping that can survive basic filters/transforms. https://deepmind.google/models/synthid/ Other providers might do something similar but I'm not aware of any. Otherwise no. There are lots of common tells, and some classifiers that might be better than guessing (at least within a given test set) but nothing reliable.", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: Is there a reliable way to tell if an image is AI generated?"}, {"id": 749754592, "item_id": 46859662, "url": "https://news.ycombinator.com/item?id=46859662", "by": "scottyah", "date": "2026-02-02", "stance": "neutral", "quote": "google and maybe others put a watermark in each image", "excerpt": "At least on image generation, google and maybe others put a watermark in each image. Text would be hard, you can't even do the printer steganography or canary traps because all models and the checker would need to have some sort of communication. https://deepmind.google/models/synthid/ You could have every provider fingerprint a message and host an API where it can attest that it's from them. I doubt the companies wo", "context": "", "use_case": "", "switched_from": "", "story": "Hacking Moltbook: The AI Social Network Any Human Can Control"}, {"id": 746813264, "item_id": 46675829, "url": "https://news.ycombinator.com/item?id=46675829", "by": "meander_water", "date": "2026-01-18", "stance": "neutral", "quote": "SynthId is used to watermark text as well, and code is open-source", "excerpt": "Not foolproof, but a couple of easy ways to verify if images were AI generated: - OpenAI uses the C2PA standard [0] to add provenance metadata to images, which you can check [1] - Gemini uses SynthId [2] and adds a watermark to the image. The watermark can be removed, but SynthId cannot as it is part of the image. SynthId is used to watermark text as well, and code is open-source [3] [0] https://help.openai.com/en/ar", "context": "", "use_case": "", "switched_from": "", "story": "Dead Internet Theory"}, {"id": 744356960, "item_id": 46522310, "url": "https://news.ycombinator.com/item?id=46522310", "by": "aprilnya", "date": "2026-01-06", "stance": "recommend", "quote": "It is, because of SynthID.", "excerpt": "It is, because of SynthID. Checked it myself: https://gemini.google.com/share/825ee3c53dd8 More info about SynthID: https://deepmind.google/models/synthid/", "context": "", "use_case": "", "switched_from": "", "story": "Debunking the AI food delivery hoax that fooled Reddit"}, {"id": 743752465, "item_id": 46484529, "url": "https://news.ycombinator.com/item?id=46484529", "by": "satvikpendem", "date": "2026-01-03", "stance": "recommend", "quote": "specific watermarking in the images themselves", "excerpt": "> And for what it's worth, the image pretty consistently trips gen AI image detectors (e.g., https://hivemoderation.com/ai-generated-content-detection). Notwithstanding your previous points, these sorts of AI detectors flag many false positives, they're not worth relying on. The only one that could actually work is specific watermarking in the images themselves, such as what Google does with SynthID in their generate", "context": "", "use_case": "", "switched_from": "", "story": "Slop Is Slop"}, {"id": 735909328, "item_id": 45994333, "url": "https://news.ycombinator.com/item?id=45994333", "by": "zaidf", "date": "2025-11-20", "stance": "compare", "quote": "", "excerpt": "This is what C2PA is trying to do: https://c2pa.org/", "context": "", "use_case": "", "switched_from": "", "story": "Nano Banana Pro"}, {"id": 735895504, "item_id": 45993469, "url": "https://news.ycombinator.com/item?id=45993469", "by": "nickdonnelly", "date": "2025-11-20", "stance": "neutral", "quote": "It's called Synth ID. It's a watermark that proves an image was generated by AI.", "excerpt": "It's called Synth ID. It's a watermark that proves an image was generated by AI. https://deepmind.google/models/synthid/", "context": "", "use_case": "", "switched_from": "", "story": "Nano Banana Pro"}]}