# c2pa Published articles for c2pa. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Authenticity Debt and the Synthetic Content Threat Landscape: A Layered Framework for Trust, Provenance, and IP Governance in the Generative AI Era DevFeed: [Authenticity Debt and the Synthetic Content Threat Landscape: A Layered Framework for Trust, Provenance, and IP Governance in the Generative AI Era]() Original publisher: [Read original article]() Author: Shubhashis Sengupta, Benjamin McCarty, Milind Savagaonkar, Rhine Andotra Published: 2026-11-03T05:00:00Z Content type: article Language: en Sources: [arXiv: Cryptography and Security]() Topics: [ai-governance](), [Generative AI](), [Zero Trust](), [human review]() Tags: [accountability](), [ai](), [ai-generated-content](), [architecture](), [attack](), [audio](), [c2pa](), [cs-ai](), [cs-cr](), [cs-cy](), [detection](), [digital-watermarking](), [enterprise](), [eu-ai-act](), [ftc](), [generated-content](), [generative](), [generative-ai](), [governance](), [human-in-the-loop](), [provenance](), [synthetic](), [trust]() ### AI overview The paper frames "authenticity debt" as institutional liability that accumulates when organizations deploy AI-generated content without preserving verifiable origin, integrity, and accountability, leaving exposure for later regulatory, legal, or market scrutiny. It surveys harms and attack vectors, along with the capabilities and failure modes of watermarking, provenance systems such as C2PA and Adobe CAI, and detection technologies. Because no single control is sufficient in open, adversarial, and changing environments, it proposes a layered approach combining cryptographic provenance, human verification, and continuous governance, informed by Zero Trust principles. The abstract also examines the EU AI Act, U.S. FTC, and NIST AI RMF. ### Source excerpt arXiv:2606.00621v2 Announce Type: replace Abstract: Generative artificial intelligence has fundamentally changed how content is now produced. It has enabled how high-fidelity text, images, audio, and videos are created, modified, and redistributed at near-zero marginal cost. This shift exposes enterprises and ecosystems to a number of risks across four reinforcing authenticity layers -- authenticity, provenance, integrity, and accountability -- that traditional controls are inadequate to address in isolation. We introduce the concept of authenticity debt: the cumulative institutional liability that accumulates when organizations deploy AI-generated content without preserving verifiable origin, integrity, and accountability, deferring exposure that surfaces under regulatory, legal, or market scrutiny. This paper presents a comprehensive, multi-dimensional taxonomy of generative AI harms and attack vectors, surveys the capabilities and failure modes of technical controls including digital watermarking, provenance frameworks (C2PA, Adobe CAI), and detection technologies, and argues that no single mechanism is sufficient in open, adversarial, and evolving environments. Drawing on Zero Trust Architecture principles and enterprise governance frameworks, we propose a layered reference architecture that integrates cryptographic provenance, human-in-the-loop verification, and continuous governance to sustain defensible authenticity at scale. We further examine the regulatory landscape (EU AI Act, U.S.\ FTC, NIST AI RMF) and identify practical guiding principles for organizations seeking to build authenticity as institutional infrastructure rather than an afterthought. ## Enhanced image forensics through partial C2PA-based provenance information DevFeed: [Enhanced image forensics through partial C2PA-based provenance information]() Original publisher: [Read original article]() Author: Riccardo Ziglio, Eleonora Marchesini, Andrea Montibeller, Giulia Boato, Silvio Ranise, Cecilia Pasquini Published: 2026-10-30T04:00:00Z Content type: article Language: en Sources: [arXiv: Artificial Intelligence](), [arXiv: Cryptography and Security]() Topics: [Digital forensics](), [Image processing](), [image editing](), [Image]() Tags: [c2pa](), [cs-ai](), [cs-cr](), [cs-mm](), [detection](), [eess-iv](), [forensics](), [image](), [provenance]() ### AI overview Partial C2PA Content Credentials record only the most recent signed operations on an image, leaving earlier provenance unknown. The paper proposes using that trusted partial history to reduce distribution shift in image forensics: first reconstruct the last signer's editing operations from the credentials' structure, then adapt a forensic detector to the image's resulting characteristics. Tests with partial credentials from commercial C2PA-enabled tools suggest the operations can be reconstructed in practical cases. In a synthetic-image detection study, this adaptation significantly improved state-of-the-art detector performance on generated or acquired images subsequently postprocessed with those tools. The evidence described is limited to this reconstruction analysis and detection case study. ### Source excerpt arXiv:2610.38639v1 Announce Type: cross Abstract: Cryptographically verifiable C2PA-based Content Credentials (CCs) can be attached to freshly acquired or generated digital images but also to already existing ones with no previous provenance information, for instance after processing through trusted C2PA-enabled software tools. The latter case leads to images with so-called partial CCs, which capture only the last operations applied, while prior provenance remains unknown and can be investigated through signal forensics. This paper addresses the forensic analysis of such images, highlighting the opportunity to leverage the partial CCs as additional trusted knowledge to adapt forensic detectors towards the statistical characteristics of the image under analysis and mitigate potential distribution shifts. To this end, a two-step strategy is envisioned: (i) reconstructing the operations applied by the last signer through a structural analysis of the partial CCs; (ii) exploiting the obtained reconstruction to enhance the forensic detector of the analyst through an adaptation process. An empirical analysis on partial CCs produced by commercial C2PA-enabled software tools through different editing operations is carried out, providing evidence on the feasibility of the reconstruction in practical cases. An experimental case study on the forensic task of synthetic image detection is then presented, showing that the proposed strategy can significantly enhance the performance of state-of-the-art detectors on test images that are postprocessed via C2PA-enabled software tools after being generated/acquired. ## Google SynthID Detector Simplifies Watermark Detection and Expands Support to Other AI Providers DevFeed: [Google SynthID Detector Simplifies Watermark Detection and Expands Support to Other AI Providers]() Original publisher: [Read original article]() Author: Sergio De Simone Published: 2026-10-21T10:00:00Z Content type: news Language: en Sources: [InfoQ]() Topics: [synthid](), [generative-ai-tools](), [Image]() Tags: [ai](), [ai-generated-content](), [ai-ml-data-engineering](), [apple](), [audio](), [c2pa](), [cryptographic](), [detection](), [development](), [facebook](), [false-positives](), [gemini](), [global](), [google](), [google-ai](), [google-deepmind](), [media](), [metadata](), [microsoft](), [ml-data-engineering](), [news](), [nvidia](), [openai](), [partners](), [synthid](), [synthid-detector](), [synthid-watermarking-detector](), [watermark]() ### AI overview SynthID Detector at SynthID.com lets users upload images, audio, or video for checks across participating SynthID partners, including OpenAI, NVIDIA, and Kakao, with Apple support expected. This consolidates detection that previously required Gemini for Google-generated content or separate provider tools. It is not a general AI detector: it covers only participating providers, not Meta or Microsoft, whose watermarking differs. A negative result means no confident SynthID signal was found, not necessarily that content is human-made; extensive edits can weaken the watermark and cause false negatives. The text also describes C2PA metadata as a complementary way to record origin and editing history, while noting that uploads or screenshots can strip it. ### Source excerpt Google DeepMind announced the global availability of SynthID Detector, a tool designed to help identify AI-generated content, including media created using AI tools from SynthID partners such as OpenAI, NVIDIA, and Kakao, with Apple support expected soon. By Sergio De Simone ## C2PA exclusion range flaw: Why signature verification must check coverage DevFeed: [C2PA exclusion range flaw: Why signature verification must check coverage]() Original publisher: [Read original article]() Author: WorkOS Published: 2026-10-19T00:00:00Z Content type: article Language: en Sources: [WorkOS]() Topics: [Cryptography](), [JSON Web Tokens](), [saml]() Tags: [attack](), [authentication](), [bug](), [c2pa](), [cryptography](), [exploit](), [federated](), [jwt](), [saml](), [verification]() ### AI overview C2PA permits byte ranges to be excluded from a signature hash. The article describes a proof of concept that excludes an entire image, allowing its signature and timestamp to verify without covering the image content. It compares this coverage gap with SAML and JWT verification failures, then recommends pinning algorithms and issuers, checking allowed coverage, reading claims only from signed bytes, and testing rejection cases. ### Source excerpt C2PA's exclusion range flaw let a signature verify while covering zero bytes. The same bug hits SAML and JWT. A checklist for verifying what a signature covers. ## OpenAI brings text watermarking to its API -- and unlike Anthropic, it's off by default DevFeed: [OpenAI brings text watermarking to its API -- and unlike Anthropic, it's off by default]() Original publisher: [Read original article]() Author: Paul Sawers Published: 2026-10-05T19:29:15Z Content type: news Language: en Sources: [The New Stack]() Topics: [watermarking](), [synthid](), [ChatGPT API](), [OpenAI]() Tags: [ai](), [ai-engineering](), [api](), [c2pa](), [claude](), [codex](), [compliance](), [eu-ai-act](), [large-language-models](), [openai](), [watermarking]() ### AI overview OpenAI now lets API customers opt in to textGrain watermarking for supported models, with the feature off by default. The article compares this choice with Anthropic's model-level watermarking and describes detection limits, including reduced performance on code and edited text. OpenAI says watermarking did not meaningfully change performance on coding and agent benchmarks, while eligibility details for automatic EU watermarking of Codex output remain unclear. ### Source excerpt OpenAI has announced that developers can now opt in to watermarking text generated through its API, as the company extends The post OpenAI brings text watermarking to its API -- and unlike Anthropic, it's off by default appeared first on The New Stack. ## The AI Economy Has a Proof Problem DevFeed: [The AI Economy Has a Proof Problem]() Original publisher: [Read original article]() Author: Elodie Aishwarya P. Remoissenet Published: 2026-10-03T08:37:31Z Content type: opinion Language: en Sources: [HackerNoon]() Topics: [Data integrity in software engineering](), [Cryptography](), [human review]() Tags: [ai](), [ai-economy](), [ai-governance](), [ai-infrastructure](), [ai-provenance](), [ai-regulation](), [c2pa](), [copilot](), [copyright-litigation](), [hackernoon-top-story]() ### AI overview The article argues that AI markets rely on labels and disclosures where verifiable evidence is missing, creating circular risks in data center financing, hiring, music provenance, and AI products that may undermine their own content supply. It proposes evidence produced during an activity, preserved as a trajectory, resilient to adversaries, and verifiable by independent parties, while acknowledging that public litigation records can remain partial and contested. ### Source excerpt We built the AI economy without an evidence layer. Litigation has become the proof layer, because nobody built one. ## Introducing Gemini 3.8 Flash TTS and Flash-Lite TTS DevFeed: [Introducing Gemini 3.8 Flash TTS and Flash-Lite TTS]() Original publisher: [Read original article]() Author: Alan Cowen Published: 2026-09-23T15:15:00Z Content type: release Language: en Sources: [Gemini Models]() Topics: [voice ai](), [speech-to-speech](), [Conversational AI](), [synthid](), [watermarking](), [Netfree]() Tags: [agent](), [ai](), [api](), [c2pa](), [content-transparency](), [developers](), [flash-lite](), [gemini-3-8-flash](), [gemini-3-8-flash-tts](), [gemini-models](), [none](), [tts]() ### AI overview Google introduces Gemini 3.8 Flash TTS for voice design and line-by-line direction, and Flash-Lite TTS for high-volume, cost-efficient speech generation. Developers can access the models through the Gemini API and Google AI Studio, with safeguards including consent verification for voice replication and SynthID audio watermarking. ### Source excerpt a text card image reading "Introducing Gemini 3.8 Flash TTS and 3.8 Flash-Lite TTS" ## AI text watermarking raises questions about future training and retrieval signals DevFeed: [AI text watermarking raises questions about future training and retrieval signals]() Original publisher: [Read original article]() Author: Pedro Dias Published: 2026-09-03T10:27:51Z Content type: article Language: en Sources: [geoSurge]() Topics: [watermarking](), [ai security](), [software dark matter]() Tags: [ai](), [ai-brand-monitoring](), [ai-brand-perception](), [ai-news](), [ai-search](), [ai-visibility](), [anthropic](), [c2pa](), [chatgpt](), [content-creation](), [content-delivery-for-geo](), [fan-out-query](), [geo](), [geo-research](), [geosurge-news](), [geosurge-research](), [geosurge-team](), [industry-analysis](), [industry-guides](), [llm-memory](), [news](), [openai](), [product](), [report](), [research](), [research-and-analysis](), [seo](), [watermark](), [watermarking](), [whitepaper]() ### AI overview Anthropic plans to watermark text generated by newer Claude models and add C2PA provenance metadata to supported file types. The article explains the transparency requirements behind the plan, compares it with Google and OpenAI efforts, and considers whether labs might eventually use provenance signals in training or retrieval. It stresses that such use has not been announced and that watermark detection can be uncertain. ### Source excerpt Anthropic will watermark text from Claude. Google has marked Gemini text since 2024. The compliance story explains the timing and nothing else. What matters is that model memory is built from a corpus nobody has ever graded - and that is changing. ## Advancing responsible AI across Europe DevFeed: [Advancing responsible AI across Europe]() Original publisher: [Read original article]() Published: 2026-07-31T15:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [responsible-ai](), [ai-governance](), [ai regulation]() Tags: [accountability](), [ai](), [ai-generated-content](), [ai-generated-media](), [ai-governance](), [audio](), [c2pa](), [compliance](), [cyber](), [eu](), [eu-ai-act](), [europe](), [global-affairs](), [governance](), [openai](), [provenance](), [responsible-ai](), [safety](), [safety-security](), [security](), [transparency]() ### AI overview OpenAI describes governance practices it says support the EU AI Act, including pre-release model testing, system cards, external red teaming, a public Model Spec, and risk frameworks for assessment, safeguards, reporting, security, and incident response. For AI-generated media provenance, it combines C2PA Content Credentials, which carry detailed context, with SynthID watermarks intended to preserve a signal when metadata is lost; it is expanding provenance to audio and working toward text coverage as standards mature. The company cautions that metadata and labels may not persist and no signal is perfect, so it supports layered measures and ecosystem cooperation. For cybersecurity, its Trusted Access for Cyber program aims to give legitimate defenders access to advanced capabilities while reducing misuse, including work with European agencies and infrastructure operators. It says it will keep updating customer resources as implementation evolves. ### Source excerpt OpenAI shares how its safety, security, transparency, and provenance practices support responsible AI governance in Europe. The work will continue as the EU AI Act advances. ## Supporting Europe's work in ensuring a trustworthy AI ecosystem DevFeed: [Supporting Europe's work in ensuring a trustworthy AI ecosystem]() Original publisher: [Read original article]() Published: 2026-06-11T00:00:00Z Content type: news Language: en Sources: [OpenAI News]() Topics: [Digital Trust](), [ai regulation](), [watermarking](), [OpenAI](), [generative-ai-tools]() Tags: [2025](), [ai](), [ai-generated-content](), [c2pa](), [chatgpt](), [code](), [codex](), [content-transparency](), [detection](), [disinformation](), [ecosystem](), [eu](), [europe](), [global-affairs](), [openai](), [provenance](), [standards](), [tools](), [transparency](), [trustworthy-ai](), [understand]() ### AI overview OpenAI supports the EU Code of Practice on Transparency of AI-Generated Content and says it will meet requirements that apply to its relevant products. Its provenance approach combines C2PA metadata, SynthID watermarks on images generated with ChatGPT, Codex, and the OpenAI API, and a public verification page for checking supported images. Metadata can carry information about a content's origin, creation or editing, and signer; watermarks can help preserve a signal across contexts. OpenAI also cites work on open standards and product safeguards such as classifiers, reporting, and enforcement. It cautions that metadata can be stripped or lost during transformations, watermarks can degrade, and labels help only where encountered. Provenance remains an emerging field requiring more reliable, robust, interoperable methods and ecosystem cooperation. ### Source excerpt OpenAI supports the EU Code of Practice on AI content transparency, advancing provenance standards and tools to help people understand AI-generated content. ## Making it easier to understand how content was created and edited DevFeed: [Making it easier to understand how content was created and edited]() Original publisher: [Read original article]() Author: Laurie Richardson Published: 2026-05-17T13:43:50Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [generative-ai-tools](), [ACR](), [Google Search](), [Google Chrome](), [Instagram]() Tags: [ai](), [ai-content](), [ai-content-detection](), [ai-generated-media](), [ai-generated-video](), [api](), [audio](), [c2pa](), [chrome](), [cloud](), [content](), [content-transparency](), [cosmos](), [digital-watermarking](), [enterprise](), [gemini](), [gemini-app](), [gemini-enterprise-agent-platform](), [gemini-in-chrome](), [generative](), [none]() ### AI overview Google says it is expanding ways to check how online media was created or edited. SynthID, which embeds imperceptible watermarks in AI-generated media, has been used on over 100 billion images and videos and 60,000 years of audio; verification is available in Gemini and is coming to Search and Chrome. C2PA Content Credentials can show how media was created or modified, including whether it is an unaltered camera original. Credential verification is rolling out in Gemini, with Search and Chrome planned later; Pixel 8, 9 and 10 are also slated to add video credentials. Google is launching a Cloud AI Content Detection API for businesses to identify media from Google and other popular models, initially with trusted partners. Adoption depends on wider industry participation: OpenAI, Kakao and ElevenLabs plan to use SynthID, while Meta plans to label camera-captured media with Content Credentials on Instagram. The announcement describes a complex, ongoing effort, not a guarantee that every item online can be verified. ### Source excerpt We're expanding our tools to help you understand how content was created and edited across the web. ## Creating with Sora Safely DevFeed: [Creating with Sora Safely]() Original publisher: [Read original article]() Published: 2026-03-23T00:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [ai safety](), [human review](), [Image]() Tags: [ai](), [ai-content](), [appearance](), [audio](), [c2pa](), [consent](), [control](), [filter](), [generate](), [guardrails](), [human-review](), [image](), [image-generation](), [image-to-video](), [impact](), [model](), [safety](), [social](), [video]() ### AI overview As of April 26, 2026, Sora is no longer available. The text describes its safety approach: videos carry visible or invisible provenance signals, including C2PA metadata, and some have moving creator watermarks. Uploading photos of people requires users to attest to consent and upload rights; generations involving people, especially children and young-looking people, receive stricter safeguards. Sora Characters let people control access to their appearance and voice, revoke permission, and review or remove videos featuring them. Other protections include teen content and messaging limits, prompt and output checks across video frames and audio transcripts, feed filtering and human review for high-impact harms, and controls to share, remove, report, or block content. The text acknowledges that audio protections are not perfect. ### Source excerpt To address the novel safety challenges posed by a state-of-the-art video model as well as a new social creation platform, we've built Sora 2 and the Sora app with safety at the foundation. Our approach is anchored in concrete protections. ## How we're bringing AI image verification to the Gemini app DevFeed: [How we're bringing AI image verification to the Gemini app]() Original publisher: [Read original article]() Author: Pushmeet Kohli Published: 2025-11-20T15:13:19Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [synthid](), [Image](), [generative-ai-tools](), [image editing]() Tags: [ai](), [ai-generated-content](), [ai-images](), [audio](), [c2pa](), [content](), [content-transparency](), [credentials](), [digital-watermarking](), [embedded](), [gemini](), [gemini-app](), [google](), [google-ai](), [google-deepmind](), [google-search](), [image](), [images](), [none](), [verification](), [verify]() ### AI overview Gemini can check whether an uploaded image carries Google's SynthID watermark. Users upload an image and ask whether it was created with Google AI; Gemini checks the watermark and uses its own reasoning to provide context. This verifies a signal associated with Google AI, not the image's full origin or history. Google says it plans to extend SynthID checks beyond images to video and audio and to more surfaces, and eventually support C2PA credentials for content from outside Google's ecosystem. Separately, images generated by Nano Banana Pro in Gemini, Vertex AI and Google Ads are beginning to include C2PA metadata. ### Source excerpt Our new Gemini app feature allows you to verify Google AI images and determine whether content was created or edited by AI. ## Stability AI's Annual Integrity Transparency Report DevFeed: [Stability AI's Annual Integrity Transparency Report]() Original publisher: [Read original article]() Author: Guest User Published: 2025-09-17T17:26:52Z Content type: article Language: en Sources: [News & Updates - Stability AI]() Topics: [responsible-ai](), [Generative AI](), [ai regulation]() Tags: [ai](), [allows](), [c2pa](), [child-safety](), [compliance](), [content-authenticity](), [report](), [transparency]() ### AI overview Stability AI's annual report describes its safety practices for generative AI models and API products during April 2024-April 2025. It covers training-data screening, API and model safeguards, red teaming, human moderation and appeals, child-safety reporting, and C2PA provenance metadata. The report also outlines planned work on risk management, audits, transparency, and oversight. ### Source excerpt At Stability AI, we are committed to building and deploying generative AI responsibly, and we believe that transparency is foundational to safe and ethical AI. ## How Pixel and Android are bringing a new level of trust to your images with C2PA Content Credentials DevFeed: [How Pixel and Android are bringing a new level of trust to your images with C2PA Content Credentials]() Original publisher: [Read original article]() Author: Edward Fernandez (noreply@blogger.com) Published: 2025-09-10T15:59:00Z Content type: article Language: en Sources: [Google Online Security]() Topics: [trust](), [Android Security](), [Android](), [Image](), [Generative AI]() Tags: [ai](), [android](), [android-security](), [c2pa](), [content](), [credentials](), [digital-media](), [images](), [pixel](), [provenance](), [trust]() ### AI overview Pixel 10 supports C2PA Content Credentials across every photo taken with Pixel Camera, and Google Photos supports them too. Google says Pixel Camera achieved C2PA Assurance Level 2, the program's highest defined rating, which it says is currently possible for a mobile app only on Android. The design uses private-by-design certificate management so images cannot be linked to one another or their creator, and on-device trusted timestamps let photos remain verifiable after certificates expire, including photos captured offline. Google attributes these capabilities to Tensor G5, Titan M2, Android's hardware-backed security, and Pixel engineering. The excerpt frames Content Credentials as a way to verify digital media's source and history using digital signatures; it does not include the promised developer implementation details. ### Source excerpt Posted by Eric Lynch, Senior Product Manager, Android Security, and Sherif Hanna, Group Product Manager, Google C2PA Core At Made by Google 2025, we announced that the new Google Pixel 10 phones will support C2PA Content Credentials in Pixel Camera and Google Photos. This announcement represents a series of steps towards greater digital media transparency: The Pixel 10 lineup is the first to have Content Credentials built in across every photo created by Pixel Camera. The Pixel Camera app achieved Assurance Level 2, the highest security rating currently defined by the C2PA Conformance Program. Assurance Level 2 for a mobile app is currently only possible on the Android platform. A private-by-design approach to C2PA certificate management, where no image or group of images can be related to one another or the person who created them. Pixel 10 phones support on-device trusted time-stamps, which ensures images captured with your native camera app can be trusted after the certificate expires, even if they were captured when your device was offline. These capabilities are powered by Google Tensor G5, Titan M2 security chip, the advanced hardware-backed security features of the Android platform, and Pixel engineering expertise. In this post, we'll break down our architectural blueprint for bringing a new level of trust to digital media, and how developers can apply this model to their own apps on Android. A New Approach to Content Credentials Generative AI can help us all to be more creative, productive, and innovative. But it can be hard to tell the difference between content that's been AI-generated, and content created without AI. The ability to verify the source and history--or provenance--of digital content is more important than ever. Content Credentials convey a rich set of information about how media such as images, videos, or audio files were made, protected by the same digital signature technology that has secured online transactions and mobile apps for decades. I