#watermark — Public Fediverse posts

I am not fussed either way that LLM output is watermarked, but I was curious how they were doing it. It seems like it's going to take a lot more compute per token. The different iterations they seemingly went through is interesting and synth-id is quite clever: https://www.youtube.com/watch?v=Cmi-1QSaptA

#ai #machinelearning #statistics #maths #watermark

I am not fussed either way that LLM output is watermarked, but I was curious how they were doing it. It seems like it's going to take a lot more compute per token. The different iterations they seemingly went through is interesting and synth-id is quite clever: https://www.youtube.com/watch?v=Cmi-1QSaptA

#ai #machinelearning #statistics #maths #watermark

Today i announced that my new watermark Backfire, in provcheck.ai 1.3.0 release survives both scrubbers  behind the published Zhao et al. paper and in Guillaume Meyer's watermarks-remover.  #AI #Watermark #EUAIACT

Today i announced that my new watermark Backfire, in provcheck.ai 1.3.0 release survives both scrubbers  behind the published Zhao et al. paper and in Guillaume Meyer's watermarks-remover.  #AI #Watermark #EUAIACT

Today i announced that my new watermark Backfire, in provcheck.ai 1.3.0 release survives both scrubbers  behind the published Zhao et al. paper and in Guillaume Meyer's watermarks-remover.  #AI #Watermark #EUAIACT

Today i announced that my new watermark Backfire, in provcheck.ai 1.3.0 release survives both scrubbers  behind the published Zhao et al. paper and in Guillaume Meyer's watermarks-remover.  #AI #Watermark #EUAIACT

Today i announced that my new watermark Backfire, in provcheck.ai 1.3.0 release survives both scrubbers  behind the published Zhao et al. paper and in Guillaume Meyer's watermarks-remover.  #AI #Watermark #EUAIACT

Почему ML-модель получает неправильные признаки: Event Time, Watermarks и Apache Flink

Представьте: вы разрабатываете антифрод-систему для банка. Пользователь совершает пять покупок подряд за пару минут. Все транзакции успешно попадают в Kafka, никаких ошибок, consumer lag почти нулевой, Flink работает стабильно. Но модель выдачи решений получает на вход признак: transactions_in_10_minute_window = 3 Хотя по факту их было пять, а вместо вероятности мошенничества 0.91 она выдаёт 0.42 .

https://habr.com/ru/articles/1074822/

#ml #mlops #java #продакшен #apache_flink #data_engineering #Watermark #streaming #realtime_ML #ai

Почему ML-модель получает неправильные признаки: Event Time, Watermarks и Apache Flink

Представьте: вы разрабатываете антифрод-систему для банка. Пользователь совершает пять покупок подряд за пару минут. Все транзакции успешно попадают в Kafka, никаких ошибок, consumer lag почти нулевой, Flink работает стабильно. Но модель выдачи решений получает на вход признак: transactions_in_10_minute_window = 3 Хотя по факту их было пять, а вместо вероятности мошенничества 0.91 она выдаёт 0.42 .

https://habr.com/ru/articles/1074822/

#ml #mlops #java #продакшен #apache_flink #data_engineering #Watermark #streaming #realtime_ML #ai

Почему ML-модель получает неправильные признаки: Event Time, Watermarks и Apache Flink

Представьте: вы разрабатываете антифрод-систему для банка. Пользователь совершает пять покупок подряд за пару минут. Все транзакции успешно попадают в Kafka, никаких ошибок, consumer lag почти нулевой, Flink работает стабильно. Но модель выдачи решений получает на вход признак: transactions_in_10_minute_window = 3 Хотя по факту их было пять, а вместо вероятности мошенничества 0.91 она выдаёт 0.42 .

https://habr.com/ru/articles/1074822/

#ml #mlops #java #продакшен #apache_flink #data_engineering #Watermark #streaming #realtime_ML #ai

MS Paint and Photos inivisibly watermark even locally generated output with GUID

https://xusheng.dev/posts/reversing/mspaint_invisible_watermark/main/

Comments: https://news.ycombinator.com/item?id=49421158

#HackerNews #MSPaint #Photos #Watermark #GUID #Privacy #Security #Tech

MS Paint and Photos inivisibly watermark even locally generated output with GUID

https://xusheng.dev/posts/reversing/mspaint_invisible_watermark/main/

Comments: https://news.ycombinator.com/item?id=49421158

#HackerNews #MSPaint #Photos #Watermark #GUID #Privacy #Security #Tech

MS Paint and Photos inivisibly watermark even locally generated output with GUID

https://xusheng.dev/posts/reversing/mspaint_invisible_watermark/main/

Comments: https://news.ycombinator.com/item?id=49421158

#HackerNews #MSPaint #Photos #Watermark #GUID #Privacy #Security #Tech

MS Paint and Photos inivisibly watermark even locally generated output with GUID

https://xusheng.dev/posts/reversing/mspaint_invisible_watermark/main/

Comments: https://news.ycombinator.com/item?id=49421158

#HackerNews #MSPaint #Photos #Watermark #GUID #Privacy #Security #Tech

MS Paint and Photos inivisibly watermark even locally generated output with GUID

https://xusheng.dev/posts/reversing/mspaint_invisible_watermark/main/

Comments: https://news.ycombinator.com/item?id=49421158

#HackerNews #MSPaint #Photos #Watermark #GUID #Privacy #Security #Tech

Google Gemini AI Image Watermark: How to Remove or Add It? Native Feature 🤖🖼💧 https://www.youtube.com/watch?v=NDQol-C62oc #Google #IA #Images #Watermark

Google Gemini AI Image Watermark: How to Remove or Add It? Native Feature 🤖🖼💧 https://www.youtube.com/watch?v=NDQol-C62oc #Google #IA #Images #Watermark

Google Gemini AI Image Watermark: How to Remove or Add It? Native Feature 🤖🖼💧 https://www.youtube.com/watch?v=NDQol-C62oc #Google #IA #Images #Watermark

Google Gemini AI Image Watermark: How to Remove or Add It? Native Feature 🤖🖼💧 https://www.youtube.com/watch?v=NDQol-C62oc #Google #IA #Images #Watermark

AI firms are watermarking generated text – here’s why it won’t work.

AI companies have begun embedding watermarks in the output of their models to improve transparency and help crack down on misinformation, disinformation and cheating on homework, but there are limitations to this approach.

https://mediafaro.org/article/20260820-ai-firms-are-watermarking-generated-text-heres-why-it-wont-work?mf_channel=mastodon&action=forward

#AI #Watermark #Tech

AI firms are watermarking generated text – here’s why it won’t work.

AI companies have begun embedding watermarks in the output of their models to improve transparency and help crack down on misinformation, disinformation and cheating on homework, but there are limitations to this approach.

https://mediafaro.org/article/20260820-ai-firms-are-watermarking-generated-text-heres-why-it-wont-work?mf_channel=mastodon&action=forward

#AI #Watermark #Tech

AI firms are watermarking generated text – here’s why it won’t work.

AI companies have begun embedding watermarks in the output of their models to improve transparency and help crack down on misinformation, disinformation and cheating on homework, but there are limitations to this approach.

https://mediafaro.org/article/20260820-ai-firms-are-watermarking-generated-text-heres-why-it-wont-work?mf_channel=mastodon&action=forward

#AI #Watermark #Tech

AI firms are watermarking generated text – here’s why it won’t work.

AI companies have begun embedding watermarks in the output of their models to improve transparency and help crack down on misinformation, disinformation and cheating on homework, but there are limitations to this approach.

https://mediafaro.org/article/20260820-ai-firms-are-watermarking-generated-text-heres-why-it-wont-work?mf_channel=mastodon&action=forward

#AI #Watermark #Tech

AI firms are watermarking generated text – here’s why it won’t work.

AI companies have begun embedding watermarks in the output of their models to improve transparency and help crack down on misinformation, disinformation and cheating on homework, but there are limitations to this approach.

https://mediafaro.org/article/20260820-ai-firms-are-watermarking-generated-text-heres-why-it-wont-work?mf_channel=mastodon&action=forward

#AI #Watermark #Tech

Reto Vogt @revogt malt sich aus, was passiert, wenn KI-Firmen gleichzeitig Täter, Forensik und Richter werden.
#AI #Watermark
https://dnip.ch/2026/08/21/vogt-am-freitag-wasserzeichen/

Reto Vogt @revogt malt sich aus, was passiert, wenn KI-Firmen gleichzeitig Täter, Forensik und Richter werden.
#AI #Watermark
https://dnip.ch/2026/08/21/vogt-am-freitag-wasserzeichen/

Reto Vogt @revogt malt sich aus, was passiert, wenn KI-Firmen gleichzeitig Täter, Forensik und Richter werden.
#AI #Watermark
https://dnip.ch/2026/08/21/vogt-am-freitag-wasserzeichen/

Reto Vogt @revogt malt sich aus, was passiert, wenn KI-Firmen gleichzeitig Täter, Forensik und Richter werden.
#AI #Watermark
https://dnip.ch/2026/08/21/vogt-am-freitag-wasserzeichen/

Reto Vogt @revogt malt sich aus, was passiert, wenn KI-Firmen gleichzeitig Täter, Forensik und Richter werden.
#AI #Watermark
https://dnip.ch/2026/08/21/vogt-am-freitag-wasserzeichen/

Last week someone shared with me the watermarks-remover GH repo. Working with C2Pa on just about every product I have this came as a fun challenge.

How do I build a watermark that not only resists stripping attempts but makes them backfire on the stripper?

Challenge accepted, received and closed.

Introducing Provcheck Backfire

https://provcheck.ai/backfire/

#AI #watermark #EUAIACT #provenance

Last week someone shared with me the watermarks-remover GH repo. Working with C2Pa on just about every product I have this came as a fun challenge.

How do I build a watermark that not only resists stripping attempts but makes them backfire on the stripper?

Challenge accepted, received and closed.

Introducing Provcheck Backfire

https://provcheck.ai/backfire/

#AI #watermark #EUAIACT #provenance

Last week someone shared with me the watermarks-remover GH repo. Working with C2Pa on just about every product I have this came as a fun challenge.

How do I build a watermark that not only resists stripping attempts but makes them backfire on the stripper?

Challenge accepted, received and closed.

Introducing Provcheck Backfire

https://provcheck.ai/backfire/

#AI #watermark #EUAIACT #provenance

Last week someone shared with me the watermarks-remover GH repo. Working with C2Pa on just about every product I have this came as a fun challenge.

How do I build a watermark that not only resists stripping attempts but makes them backfire on the stripper?

Challenge accepted, received and closed.

Introducing Provcheck Backfire

https://provcheck.ai/backfire/

#AI #watermark #EUAIACT #provenance

Last week someone shared with me the watermarks-remover GH repo. Working with C2Pa on just about every product I have this came as a fun challenge.

How do I build a watermark that not only resists stripping attempts but makes them backfire on the stripper?

Challenge accepted, received and closed.

Introducing Provcheck Backfire

https://provcheck.ai/backfire/

#AI #watermark #EUAIACT #provenance

Невидимые метки в AI-картинках 2.0: JPEG, blur, перерисовка и ошибки первого теста

После первой статьи в комментариях попросили больше технической части и меньше выводов по одной картинке. Разбираюсь, что SynthID переживает, что может разрушить метку и что вообще доказывает её наличие. В первой статье я проверил одно изображение через сторонний сервис и сделал из этого слишком широкий вывод. В комментариях предложили более нормальный набор тестов: JPEG-сжатие, resize, crop, blur, скриншот, повторную генерацию и анализ частотной области. Там же мне скинули reverse engineering SynthID. Так что здесь разбираемся уже не с одним сервисом, а с самой технологией. Услышал ваши комментарии, исправляюсь

https://habr.com/ru/companies/syntx_ai/articles/1072542/

#SynthID #невидимые_водяные_знаки #AIизображения #watermark #JPEG #обработка_изображений #DeepMind #OpenAI_Verify #C2PA #генеративный_ИИ

Невидимые метки в AI-картинках 2.0: JPEG, blur, перерисовка и ошибки первого теста

После первой статьи в комментариях попросили больше технической части и меньше выводов по одной картинке. Разбираюсь, что SynthID переживает, что может разрушить метку и что вообще доказывает её наличие. В первой статье я проверил одно изображение через сторонний сервис и сделал из этого слишком широкий вывод. В комментариях предложили более нормальный набор тестов: JPEG-сжатие, resize, crop, blur, скриншот, повторную генерацию и анализ частотной области. Там же мне скинули reverse engineering SynthID. Так что здесь разбираемся уже не с одним сервисом, а с самой технологией. Услышал ваши комментарии, исправляюсь

https://habr.com/ru/companies/syntx_ai/articles/1072542/

#SynthID #невидимые_водяные_знаки #AIизображения #watermark #JPEG #обработка_изображений #DeepMind #OpenAI_Verify #C2PA #генеративный_ИИ

Невидимые метки в AI-картинках 2.0: JPEG, blur, перерисовка и ошибки первого теста

После первой статьи в комментариях попросили больше технической части и меньше выводов по одной картинке. Разбираюсь, что SynthID переживает, что может разрушить метку и что вообще доказывает её наличие. В первой статье я проверил одно изображение через сторонний сервис и сделал из этого слишком широкий вывод. В комментариях предложили более нормальный набор тестов: JPEG-сжатие, resize, crop, blur, скриншот, повторную генерацию и анализ частотной области. Там же мне скинули reverse engineering SynthID. Так что здесь разбираемся уже не с одним сервисом, а с самой технологией. Услышал ваши комментарии, исправляюсь

https://habr.com/ru/companies/syntx_ai/articles/1072542/

#SynthID #невидимые_водяные_знаки #AIизображения #watermark #JPEG #обработка_изображений #DeepMind #OpenAI_Verify #C2PA #генеративный_ИИ

The flawed detection strategy has simply been formalized. Attractors are not new.

They are cliche.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

#ai #writing #watermark

The flawed detection strategy has simply been formalized. Attractors are not new.

They are cliche.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

#ai #writing #watermark

The flawed detection strategy has simply been formalized. Attractors are not new.

They are cliche.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

#ai #writing #watermark

The flawed detection strategy has simply been formalized. Attractors are not new.

They are cliche.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

#ai #writing #watermark

The flawed detection strategy has simply been formalized. Attractors are not new.

They are cliche.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

#ai #writing #watermark

Claudes unsichtbares Wasserzeichen: So wirst du es los

https://www.youtube.com/watch?v=w25dcxZvouU

#claude #watermark #anthropic #euaiact

Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing

Earlier this month, Anthropic announced that future versions of Claude will generate text that includes watermarks showing it was AI-generated. At the time, Anthropic did not explain how this would work, leaving us to speculate on the podcast: Would it somehow encode this into the text? Include invisible characters? Do something with the metadata? We now know, thanks to a blog post over the weekend, that Anthropic will do this by changing how its AI writes altogether.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing

Earlier this month, Anthropic announced that future versions of Claude will generate text that includes watermarks showing it was AI-generated. At the time, Anthropic did not explain how this would work, leaving us to speculate on the podcast: Would it somehow encode this into the text? Include invisible characters? Do something with the metadata? We now know, thanks to a blog post over the weekend, that Anthropic will do this by changing how its AI writes altogether.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing

Earlier this month, Anthropic announced that future versions of Claude will generate text that includes watermarks showing it was AI-generated. At the time, Anthropic did not explain how this would work, leaving us to speculate on the podcast: Would it somehow encode this into the text? Include invisible characters? Do something with the metadata? We now know, thanks to a blog post over the weekend, that Anthropic will do this by changing how its AI writes altogether.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing

Earlier this month, Anthropic announced that future versions of Claude will generate text that includes watermarks showing it was AI-generated. At the time, Anthropic did not explain how this would work, leaving us to speculate on the podcast: Would it somehow encode this into the text? Include invisible characters? Do something with the metadata? We now know, thanks to a blog post over the weekend, that Anthropic will do this by changing how its AI writes altogether.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing

Earlier this month, Anthropic announced that future versions of Claude will generate text that includes watermarks showing it was AI-generated. At the time, Anthropic did not explain how this would work, leaving us to speculate on the podcast: Would it somehow encode this into the text? Include invisible characters? Do something with the metadata? We now know, thanks to a blog post over the weekend, that Anthropic will do this by changing how its AI writes altogether.

https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing/

How AI text watermarking works: a visual guide

"A gentle visual guide to how statistical watermarks hide inside AI-generated text, and what erases them."

Link: https://declaude.org/watermarking/

#linkdump #ai #watermark

How AI text watermarking works: a visual guide

"A gentle visual guide to how statistical watermarks hide inside AI-generated text, and what erases them."

Link: https://declaude.org/watermarking/

#linkdump #ai #watermark

How AI text watermarking works: a visual guide

"A gentle visual guide to how statistical watermarks hide inside AI-generated text, and what erases them."

Link: https://declaude.org/watermarking/

#linkdump #ai #watermark

How AI text watermarking works: a visual guide

"A gentle visual guide to how statistical watermarks hide inside AI-generated text, and what erases them."

Link: https://declaude.org/watermarking/

#linkdump #ai #watermark

How AI text watermarking works: a visual guide

"A gentle visual guide to how statistical watermarks hide inside AI-generated text, and what erases them."

Link: https://declaude.org/watermarking/

#linkdump #ai #watermark

Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing

"It’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance.

Link: https://daringfireball.net/2026/08/anthropics_watermark_text_adulteration_in_claude_is_a_perversion_of_writing

#linkdump #ai #blogpost #claude #watermark

Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing

"It’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance.

Link: https://daringfireball.net/2026/08/anthropics_watermark_text_adulteration_in_claude_is_a_perversion_of_writing

#linkdump #ai #blogpost #claude #watermark

Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing

"It’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance.

Link: https://daringfireball.net/2026/08/anthropics_watermark_text_adulteration_in_claude_is_a_perversion_of_writing

#linkdump #ai #blogpost #claude #watermark

Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing

"It’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance.

Link: https://daringfireball.net/2026/08/anthropics_watermark_text_adulteration_in_claude_is_a_perversion_of_writing

#linkdump #ai #blogpost #claude #watermark

Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing

"It’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance.

Link: https://daringfireball.net/2026/08/anthropics_watermark_text_adulteration_in_claude_is_a_perversion_of_writing

#linkdump #ai #blogpost #claude #watermark

#Anthropic #watermark : "…a textbook example (I plan to use it in my classes!) to explain why the “ #techpolicy moves slowly because politicians don’t understand tech” narrative is ridiculous. Tech policy does move slowly, but for the same reasons all policy moves slowly. Figuring out the right #tradeoffs in almost any situation — and getting public buy-in — is genuinely hard and can never happen at the speed of tech. If tech policy were to move even half as fast as people constantly claim they want it to move, absolute chaos would result. This situation is a good example."

https://x.com/random_walker/status/2089414077286166911?s=46

#Anthropic #watermark : "…a textbook example (I plan to use it in my classes!) to explain why the “ #techpolicy moves slowly because politicians don’t understand tech” narrative is ridiculous. Tech policy does move slowly, but for the same reasons all policy moves slowly. Figuring out the right #tradeoffs in almost any situation — and getting public buy-in — is genuinely hard and can never happen at the speed of tech. If tech policy were to move even half as fast as people constantly claim they want it to move, absolute chaos would result. This situation is a good example."

https://x.com/random_walker/status/2089414077286166911?s=46