#anthropic — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #anthropic, aggregated by home.social.
-
Anthropic will add machine-readable digital labels to text and files created with new Claude models released in the EU after Aug. 2, and says the system will apply worldwide. Even human-written material that Claude only proofreads, formats or translates may be marked. https://sfg.media/en/a/anthropic-claude-machine-readable-digital-labels #Anthropic #Claude #AI #EUAIAct
-
Anthropic will add machine-readable digital labels to text and files created with new Claude models released in the EU after Aug. 2, and says the system will apply worldwide. Even human-written material that Claude only proofreads, formats or translates may be marked. https://sfg.media/en/a/anthropic-claude-machine-readable-digital-labels #Anthropic #Claude #AI #EUAIAct
-
Weltweite Einführung: Wasserzeichen für mit Claude erstellte Texte https://www.computerbase.de/news/apps/weltweite-einfuehrung-anthropic-versieht-mit-claude-erstellte-texte-mit-wasserzeichen.98812/ #anthropic #claude
-
Weltweite Einführung: Wasserzeichen für mit Claude erstellte Texte https://www.computerbase.de/news/apps/weltweite-einfuehrung-anthropic-versieht-mit-claude-erstellte-texte-mit-wasserzeichen.98812/ #anthropic #claude
-
Was glaubt ihr ist zuerst online? Die neuen Wasserzeichen von #Anthropic oder ein Tool mit dem man diese entfernen kann? Wird sicherlich knapp!
-
Was glaubt ihr ist zuerst online? Die neuen Wasserzeichen von #Anthropic oder ein Tool mit dem man diese entfernen kann? Wird sicherlich knapp!
-
Google e Anthropic desenvolveram uma técnica para aplicar marcas de água invisíveis em textos, utilizando a ordem das palavras como selo de autoria. Esta técnica permite identificar conteúdo gerado por inteligência artificial de forma mais eficaz." 🤖
-
Google e Anthropic desenvolveram uma técnica para aplicar marcas de água invisíveis em textos, utilizando a ordem das palavras como selo de autoria. Esta técnica permite identificar conteúdo gerado por inteligência artificial de forma mais eficaz." 🤖
-
#KI setzt unsichtbares #Wasserzeichen in Texte und Dateien.
Wird hier dem Schummeln vorgebeugt, oder geistiges Eigentum, das mit Hilfe von KI formuliert wurde, umgelabelt?
Die Diskussion wird jedenfalls knallhart und Antrophic kann sich schon einmal warm anziehen ob ihres informationsgebaren gegenüber seinen Kunden.
KI-Texterkennung: #Anthropic versieht Claude mit unsichtbarem Wasserzeichen seit den 2.8.2026. breit kommuniziert wird es seit dem 11.08.2026.
https://the-decoder.de/ki-texterkennung-anthropic-versieht-claude-mit-unsichtbarem-wasserzeichen/
-
Claude advances the Riemann hypothesis from 41.6% to 67.2% verified zeros, OpenAI tests ChatGPT ads, and hidden AI reasoning traces exposed via jailbreak. A two-month startup raises $1.1B.
-
Claude advances the Riemann hypothesis from 41.6% to 67.2% verified zeros, OpenAI tests ChatGPT ads, and hidden AI reasoning traces exposed via jailbreak. A two-month startup raises $1.1B.
-
My favorite Fable 5 report, the 7th edition:
Conjecture ledger and primes graph:Fable 5 - 7th edition ➡️ https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
For privacy ➡️ https://web.archive.org/web/20260806034509/https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
#decompwlj #math #AI #Anthropic #Claude #Fable #Fable5 #mathematics #OEIS #sequence #numbers #graph #PrimeNumbers #NumberTheory #sieve #arithmetic #threejs #research #sieveOfEratosthenes
-
My favorite Fable 5 report, the 7th edition:
Conjecture ledger and primes graph:Fable 5 - 7th edition ➡️ https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
For privacy ➡️ https://web.archive.org/web/20260806034509/https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
#decompwlj #math #AI #Anthropic #Claude #Fable #Fable5 #mathematics #OEIS #sequence #numbers #graph #PrimeNumbers #NumberTheory #sieve #arithmetic #threejs #research #sieveOfEratosthenes
-
#Anthropic integriert in seine #Claude-Modelle ein unsichtbares Wasserzeichen zur Kennzeichnung von #KI-generierten Inhalten. Das soll Transparenz schaffen und das unerkannte Weitergeben als eigene Leistung erschweren. Das Wasserzeichen beeinträchtigt weder Lesbarkeit noch Bedeutung der Texte. https://www.heise.de/news/Keine-Chance-fuer-Schummler-Claude-bekommt-jetzt-ein-unsichtbares-Wasserzeichen-11410129.html?seite=all
-
#Anthropic integriert in seine #Claude-Modelle ein unsichtbares Wasserzeichen zur Kennzeichnung von #KI-generierten Inhalten. Das soll Transparenz schaffen und das unerkannte Weitergeben als eigene Leistung erschweren. Das Wasserzeichen beeinträchtigt weder Lesbarkeit noch Bedeutung der Texte. https://www.heise.de/news/Keine-Chance-fuer-Schummler-Claude-bekommt-jetzt-ein-unsichtbares-Wasserzeichen-11410129.html?seite=all
-
Claude: KI-Wasserzeichen – ein Bärendienst für die Wahrheit | heise online https://www.heise.de/meinung/KI-Wasserzeichen-in-Text-Ein-Stempel-der-falsche-Sicherheit-schafft-11410549.html #ArtificialIntelligence #AI #AIact #ArtificialIntelligenceAct #Anthropic
-
Claude: KI-Wasserzeichen – ein Bärendienst für die Wahrheit | heise online https://www.heise.de/meinung/KI-Wasserzeichen-in-Text-Ein-Stempel-der-falsche-Sicherheit-schafft-11410549.html #ArtificialIntelligence #AI #AIact #ArtificialIntelligenceAct #Anthropic
-
#Anthropic has committed to #marking #AIgeneratedcontent from #Claude models launched on or after August 2, 2026, with #machinereadable #watermarks and provenance #metadata. This marking will apply across all Claude products and cloud partners, #worldwide. While the marks provide transparency, they are not conclusive proof of content provenance and may not be present in heavily edited or short content. https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content?eicker.news #tech #news #ainews
-
#Anthropic has committed to #marking #AIgeneratedcontent from #Claude models launched on or after August 2, 2026, with #machinereadable #watermarks and provenance #metadata. This marking will apply across all Claude products and cloud partners, #worldwide. While the marks provide transparency, they are not conclusive proof of content provenance and may not be present in heavily edited or short content. https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content?eicker.news #tech #news #ainews
-
RT @Polymarket: EILMELDUNG: Anthropic schließt einen 20-Jahres-Vertrag im Wert von 9,1 Milliarden US-Dollar mit dem Bitcoin-Miner Riot Platforms, um die Rechenkapazität für KI zu sichern.
mehr auf Arint.info
#AIInfrastruktur #Anthropic #BitcoinMining #KIKapazität #RiotPlatforms #TechNews #arint_info
-
Anthropic will watermark all Claude outputs globally using C2PA, with labels built into new models from August 2026 and third-party detection tools planned.
Source: The Decoder AI
https://the-decoder.com/anthropic-watermarks-all-claude-outputs-globally-with-marks-that-may-persist-through-some-editing/ -
Anthropic will watermark all Claude outputs globally using C2PA, with labels built into new models from August 2026 and third-party detection tools planned.
Source: The Decoder AI
https://the-decoder.com/anthropic-watermarks-all-claude-outputs-globally-with-marks-that-may-persist-through-some-editing/ -
My favorite Fable 5 report, the 7th edition:
Conjecture ledger and primes graph:Fable 5 - 7th edition ➡️ https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
For privacy ➡️ https://web.archive.org/web/20260806034509/https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
#decompwlj #math #AI #Anthropic #Claude #Fable #Fable5 #mathematics #OEIS #sequence #numbers #graph #PrimeNumbers #NumberTheory #sieve #arithmetic #threejs #research #sieveOfEratosthenes
-
My favorite Fable 5 report, the 7th edition:
Conjecture ledger and primes graph:Fable 5 - 7th edition ➡️ https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
For privacy ➡️ https://web.archive.org/web/20260806034509/https://decompwlj.com/Fable5_decompwlj_deep_analysis_7th_edition_2026_08_05.html
#decompwlj #math #AI #Anthropic #Claude #Fable #Fable5 #mathematics #OEIS #sequence #numbers #graph #PrimeNumbers #NumberTheory #sieve #arithmetic #threejs #research #sieveOfEratosthenes
-
Finally Anthropic did something anthropic
-
Finally Anthropic did something anthropic
-
OpenAI’s Brad Lightcap exits, extending a leadership exodus ahead of its planned IPO. Recent departures include the head of ethics, head of safety systems, and mission alignment chief as competition with Anthropic intensifies.
Source: Semafor Africa
https://www.semafor.com/article/08/11/2026/more-openai-executives-exit -
OpenAI’s Brad Lightcap exits, extending a leadership exodus ahead of its planned IPO. Recent departures include the head of ethics, head of safety systems, and mission alignment chief as competition with Anthropic intensifies.
Source: Semafor Africa
https://www.semafor.com/article/08/11/2026/more-openai-executives-exit -
Anthropic will embed invisible watermarks in Claude’s AI-generated text starting August 2 to meet EU AI Act rules and help identify AI-assisted content. Images will get watermarks too.
Source: Fortune Term Sheet
https://fortune.com/2026/08/11/anthropic-claude-watermark-ai-text-police-ai-slop/ -
Anthropic will embed invisible watermarks in Claude’s AI-generated text starting August 2 to meet EU AI Act rules and help identify AI-assisted content. Images will get watermarks too.
Source: Fortune Term Sheet
https://fortune.com/2026/08/11/anthropic-claude-watermark-ai-text-police-ai-slop/ -
41,6% → 67,2%. Нейросеть далеко обогнала человеческий рекорд по гипотезе Римана
В комментариях к статье про теорему Миттаг-Леффлера я пообещал разобрать гипотезу Римана понятно и подробно. Неожиданно оказалось, что пока я писал текст, ИИ произвел в этой области исследований новый прорыв и поэтому содержание статьи пришлось расширить. 10 августа 2026-го года Anthropic выложила препринт на тридцать пять страниц. В графе «автор» они указали Claude. Утверждение о достигнутом результате выглядит невероятно: ИИ доказал, что не менее двух третей нетривиальных нулей дзета-функции лежат на критической прямой. В последние десятки лет прогресс в этой задаче измерялся в сотых долях одного процента и каждый такой сдвиг считался достижением: настолько хватало мощи человеческого гения. В этой статье я расскажу о том, что это всё значит и почему это очень круто. Ниже также — обещанный вывод связи между нулями дзеты и простыми числами, потом что именно доказано, как (идея действительно умещается на страницу и она красивая), как её нашли — там настоящий детектив, — и чего всё это не значит. Последнее важнее всего, потому что новостные заголовки на эту тему врут как минимум наполовину.
https://habr.com/ru/articles/1069480/
#гипотеза_Римана #дзетафункция_Римана #теория_чисел #Claude #Anthropic #искусственный_интеллект #комплексный_анализ #математика #парная_корреляция #AGI
-
Anthropic führt Wasserzeichen bei KI-Texten ein
Anthropic will Texte seines Chatbots Claude mit einem unsichtbaren Wasserzeichen versehen. Grund sind neue EU-Vorgaben: KI-Inhalte müssen künftig mit maschinenlesbaren Markierungen versehen werden.
-
Anthropic führt Wasserzeichen bei KI-Texten ein
Anthropic will Texte seines Chatbots Claude mit einem unsichtbaren Wasserzeichen versehen. Grund sind neue EU-Vorgaben: KI-Inhalte müssen künftig mit maschinenlesbaren Markierungen versehen werden.
-
Claude text gains invisible watermarks across 5 Anthropic products: Marks travel with copied text and cover AWS, Google Cloud and Microsoft Foundry output. Detection tooling has not shipped, and older models remain unmarked. https://ppc.land/claude-text-gains-invisible-watermarks-across-5-anthropic-products/ #ClaudeAI #InvisibleWatermarks #Anthropic #TextDetection #AWS
-
Companies are purchasing and destroying books to train #LLM models
https://www.rnz.co.nz/life/books/wellington-second-hand-book-store-s-mysterious-orders
-
Companies are purchasing and destroying books to train #LLM models
https://www.rnz.co.nz/life/books/wellington-second-hand-book-store-s-mysterious-orders
-
AI watermarks in text: A stamp that creates false security
The EU demands labeling for AI content. Anthropic responds with watermarks in Claude – but the technology carries risks for users and the truth.
#AIAct #Anthropic #Claude #Datenschutz #EU #IT #KünstlicheIntelligenz #news
-
AI watermarks in text: A stamp that creates false security
The EU demands labeling for AI content. Anthropic responds with watermarks in Claude – but the technology carries risks for users and the truth.
#AIAct #Anthropic #Claude #Datenschutz #EU #IT #KünstlicheIntelligenz #news
-
OpenAI: Head of AI Ethics Departs After Less Than a Year
OpenAI loses its head of AI ethics. Meanwhile, former robotics head Caitlin Kalinowski is now working for Anthropic.
-
OpenAI: Head of AI Ethics Departs After Less Than a Year
OpenAI loses its head of AI ethics. Meanwhile, former robotics head Caitlin Kalinowski is now working for Anthropic.
-
IA: O perigo está nos “modelos abertos”?
EUA impõem limites às big techs, mas isenta “versões genéricas” e abrem debate sobre roubo de propriedade intelectual e segurança. Mas falhas envolvendo “agentes” poderosos revelam, sobretudo, a imprudência das corporações em meio a uma bolha cada vez mais infladahttps://outraspalavras.net/tecnologiaemdisputa/ia-o-perigo-esta-nos-modelos-abertos/
-
IA: O perigo está nos “modelos abertos”?
EUA impõem limites às big techs, mas isenta “versões genéricas” e abrem debate sobre roubo de propriedade intelectual e segurança. Mas falhas envolvendo “agentes” poderosos revelam, sobretudo, a imprudência das corporações em meio a uma bolha cada vez mais infladahttps://outraspalavras.net/tecnologiaemdisputa/ia-o-perigo-esta-nos-modelos-abertos/
-
A OpenAI lançou o app do ChatGPT para Linux, disponibilizando-o em pré-visualização global. A aplicação foi bem recebida pela comunidade de programadores de código aberto.
-
A OpenAI lançou o app do ChatGPT para Linux, disponibilizando-o em pré-visualização global. A aplicação foi bem recebida pela comunidade de programadores de código aberto.
-
Anthropic will enable default watermarking across all Claude products globally, including APIs and enterprise deployments. In India, MeitY has proposed binding rules to label deepfakes and AI-generated content as synthetically generated.
Source: Inc42
https://inc42.com/features/anthropic-adds-watermarks-in-claude-content-as-deepfake-bad-actors-run-amok/ -
⬆️ @kim_harding
There was once a very successful frizzonaire company particularly famous for its amazing lammour obstans.
Secret to its success was its highly technical staff trained in multiple disciplines.Through creative thinking they secured patents for their glozzom and durrsore processes that gave their lammour obstans a clear lead in the market.
One day they got rid of their human staff after giving in to pitches to SHARE THEIR KNOWHOW & outsource their work to #OpenAI, #Anthropic …🤦
-
An unreleased Anthropic model made progress on one of math’s biggest unsolved problems. Anthropic's unreleased model achieves unexpected progress on the Riemann hypothesis, a 150-year-old unsolved math problem, signaling a potential breakthrough in AI's mathematical reasoning capabilities.
Source: TechCrunch AI
https://techcrunch.com/2026/08/11/an-unreleased-anthropic-model-made-progress-on-one-of-maths-biggest-unsolved-problems/ -
Wenn ich unterwegs bin und mein #apple #macbookair benutze finde ich schon ab und an, dass mein stationärer #linuxmint Rechner noch Luft nach oben hat, was den #bedienkomfort #usability betrifft. Wenn's doch nur 'ne europäische KI gäbe, die mir als Companion für mein Linux Mint die #claudeai von #anthropic ersetzen könnte. Europas Rückstand in Sachen Technologie for costumers ist schon frustrierend.
-
Wenn ich unterwegs bin und mein #apple #macbookair benutze finde ich schon ab und an, dass mein stationärer #linuxmint Rechner noch Luft nach oben hat, was den #bedienkomfort #usability betrifft. Wenn's doch nur 'ne europäische KI gäbe, die mir als Companion für mein Linux Mint die #claudeai von #anthropic ersetzen könnte. Europas Rückstand in Sachen Technologie for costumers ist schon frustrierend.
-
كشفت Anthropic عن تضمين علامات مائية غير مرئية في النصوص التي يولدها Claude، وبيانات وصفية رقمية للصور بمعيار C2PA، امتثالاً لمتطلبات الشفافية بقانون الاتحاد الأوروبي للذكاء الاصطناعي. ستُطبق هذه العلامات عالمياً عبر جميع منتجات Claude وخدمات الطرف الثالث مثل AWS وGoogle Cloud. وتعمل الشركة على تطوير أدوات لاكتشافها، مع التأكيد على أن وجود العلامة يشير فقط إلى معالجة Claude للمحتوى، وليس بالضرورة إنشائه الأصلي، كما يمكن إزالتها أو يصعب اكتشافها في بعض الحالات.
-
Что на самом деле лежит в зашифрованном блоке рассуждений Claude: разбираем signature по байтам
Сегодня на Хабре вышла статья про кражу цепочек рассуждений у закрытых моделей. Схема такая: берёшь зашифрованный блок сильной модели, подкладываешь слабой сестре того же провайдера, просишь пересказать дословно, и читаешь чужие мысли, включая случайно попавшие туда пароли и ключи. Атака красивая и, судя по препринту, работающая. Читал я её с двойным чувством, потому что это самое поле signature , которое Claude прикладывает к каждому блоку thinking , я разбирал по байтам ещё прошлой зимой, когда ковырял историю своего агента. Тогда набралось около полутора тысяч таких блоков из локальных сессий. Вывод был простой: конверт вокруг рассуждения совсем не opaque. Это обычный protobuf, он читается вообще без ключей, и зашифрован внутри ровно один вложенный кусок, сам текст мысли. Всё остальное лежит открыто. Статья заставила меня сдуть пыль с того исследования и перегнать парсер по свежим сессиям. За полгода протокол подрос на четыре версии схемы, и внутри нашлось кое-что, из-за чего у меня к выводам препринта есть уточнение. Но давайте по порядку. Начнём с одной маленькой, настоящей подписи, которую я вытащил из собственной сессии Claude Code полчаса назад. Откроем конверт...
-
KI ist nicht mehr nur Mathe-Werkzeug, sondern wird zum Mitautor. Eine unveröffentlichte Version von Claude hat eine zentrale Grenze der Riemannschen Vermutung von 41,6 auf 67,2 Prozent verbessert. 60 Sub-Agenten schrieben Python-Skripte und suchten Gegenbeispiele. Der Beweis ist formal fehlerfrei und von Experten bestätigt. #Claude #Anthropic #RiemannscheVermutung #Mathe #AIGeneratedImage
-
Claude Will Soon Leave a Hidden Mark on Everything It Writes.
-
Claude Will Soon Leave a Hidden Mark on Everything It Writes.
-
Anthropic's AI model has made notable progress on the Riemann hypothesis, coordinating 60 sub-agents over 36 hours to test 650 approaches. This highlights AI's growing role in tackling complex mathematical problems and prompts a reevaluation of traditional methodologies in the field.
#AI #Anthropic #RiemannHypothesis #Mathematics #MachineLearning #TechNews
https://thedailytechfeed.com/anthropics-ai-model-advances-on-riemann-hypothesis/
-
Anthropic will embed watermarks in text and C2PA metadata in media outputs from Claude globally, with EU models adopting machine-readable marking by August 2026 to meet the EU AI Act's transparency requirements.
Source: MediaNama
https://www.medianama.com/2026/08/223-anthropic-watermark-c2pa-metadata-claude/