Google unveils SynthID Detector to verify AI-generated content
Google has introduced the SynthID Detector portal, a tool designed to help users verify whether content has been generated using Google AI by detecting SynthID watermarks.
The SynthID Detector provides a centralised platform where users can upload images, audio, video, or text to determine if the material contains a SynthID watermark. The portal is intended to address growing concerns around content authenticity, context, and verification as generative AI technologies continue to develop and become more widespread.
Pushmeet Kohli, Vice President of Science and Strategic Initiatives at Google DeepMind, commented, "Advances in generative AI are making it possible for people to create content in entirely new ways — from text to high quality audio, images and videos. As these capabilities advance and become more broadly available, questions of authenticity, context and verification emerge."
Kohli added, "Today we're announcing SynthID Detector, a verification portal to quickly and efficiently identify AI-generated content made with Google AI. The portal provides detection capabilities across different modalities in one place and provides essential transparency in the rapidly evolving landscape of generative media. It can also highlight which parts of the content are more likely to have been watermarked with SynthID."
The SynthID watermarking technology was initially developed to embed imperceptible watermarks in AI-generated images to help mitigate misinformation and misattribution. These watermarks are designed to remain detectable even when the content is shared or altered. Google states that over 10 billion pieces of content have already been watermarked with SynthID since the system's launch. SynthID's coverage has since expanded beyond imagery to include text, audio, and video content generated by models such as Gemini, Imagen, Lyria, and Veo.
Kohli explained, "When we launched SynthID — a state-of-the-art tool that embeds imperceptible watermarks and enables the identification of AI-generated content — our aim was to provide a suite of novel technical solutions to help minimise misinformation and misattribution. SynthID not only preserves the content's quality, it acts as a robust watermark that remains detectable even when the content is shared or undergoes a range of transformations. While originally focused on AI-generated imagery only, we've since expanded SynthID to cover AI-generated text, audio and video content, including content generated by our Gemini, Imagen, Lyria and Veo models across Google. Over 10 billion pieces of content have already been watermarked with SynthID."
To use the SynthID Detector, users upload an image, audio track, video, or text generated using Google's tools to the portal. The system scans the content for SynthID watermarks, highlighting particular segments most likely to be watermarked. For audio, specific time segments are indicated, while images are displayed with areas that likely contain the watermark.
The operational process is described as follows: "Upload content: Upload an image, audio track, video or piece of text created using Google's AI tools. Scan for watermarks: The portal scans the uploaded media and detects whether the uploaded content, or specific portions of it, contain a SynthID watermark. View results: The portal will present the results and if a SynthID watermark is detected, highlight which parts of the content are more likely to have been watermarked."
The SynthID Detector is currently being rolled out to early testers before a wider release. Google notes that journalists, media professionals, and researchers are eligible to join a waitlist for early access to the portal.
Google has also made the SynthID text watermarking framework open source. This enables developers to integrate the technology into their own models and projects. In addition to open sourcing, Google has begun to build partnerships with other companies to expand the reach and adoption of SynthID.
Earlier this year, Google entered a partnership with NVIDIA to embed SynthID watermarks into videos generated by NVIDIA's CosmosTM preview NIM microservice. This collaboration increases the volume of content across the internet bearing the SynthID marker, broadening transparency beyond content produced solely by Google's AI models.
Google has also announced a new partnership with GetReal Security, a company specialising in content verification, to further the capabilities of SynthID detection by external parties. The partnership is intended to support the detection of SynthID watermarks across a wider range of media and stakeholders.
On the challenge of content transparency in the context of generative AI, Kohli stated, "Content transparency remains a complex challenge. To continue to inform and empower people engaging with AI-generated content, we believe it's vital to continue collaborating with the AI community and broaden access to transparency tools."
Kohli also acknowledged the contributions of Miguel de Andrés-Clavera and the Google Partner Innovation team in developing the SynthID Detector portal.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Techday NZ
21 minutes ago
- Techday NZ
Experts warn of surge in Google, Apple, Microsoft breaches
Cybersecurity experts are raising alarm over a significant campaign targeting users through the Google Chrome Web Store, as well as the discovery of a vast database containing hundreds of millions of stolen log-in credentials. The recent developments underscore rising risks associated with browser extensions and the continuing vulnerabilities in digital identity platforms. "A Google Chrome Web Store campaign is using over 100 malicious browsers that mimic tools like VPNs, AI assistants, and crypto utilities to steal cookies and execute remote scripts secretly. Though Google has removed many extensions identified, some still remain on the Web Store," said Andrew Costis, Engineering Manager of the Adversary Research Team at AttackIQ. "The campaign relies on malvertising strategies to trick users into clicking buttons that link to malicious browser extensions. The extensions connect the victim to the threat actor's infrastructure, allowing information to be stolen, as well as modifying network traffic to deliver ads, perform redirections, or serve as a proxy. "With some of these extensions still active on the Chrome Web Store, it is essential that individuals and organizations take appropriate precautions. Knowledge is key -- users should only trust proven, reputable publishers and familiarize themselves with lure website domains. Additionally, organizations should implement adversarial exposure validation tools to ensure their security systems are tested against malicious browser campaigns." The campaign's persistence highlights the challenges facing platform operators like Google in completely eradicating malicious content from widely used app stores. With new extensions and techniques emerging regularly, the risk to end users remains ongoing. Meanwhile, cybersecurity concerns have been exacerbated by the discovery of a database containing an estimated 184 million records of stolen log-in credentials. The database reportedly contains detailed access information for popular services, including Apple, Microsoft, Google, Facebook, Instagram, Snapchat, as well as various banking, healthcare, and government platforms across numerous countries. "What's most noteworthy is how this breach highlights the immense value of centralized identity platforms like Google, Okta, Apple and Meta to attackers. With over 184 million records exposed, threat actors can now launch widespread account takeover attempts across countless SaaS applications and cloud services that rely on these providers for authentication," sid Cory Michal, Chief Security Officer at AppOmni. "This is not surprising. Databases like this are regularly bought, sold, and repackaged on dark web forums like BreachForums. Massive credential dumps are part of an ongoing black market where breached data is commoditized and often aggregated from multiple incidents over time. What's new isn't the existence of the data, but the scale, the recency of some credentials, and the targeting of identity providers that are widely used to access SaaS and cloud services—making this breach especially potent for enabling downstream account takeovers. "This breach calls attention to a bigger issue. We increasingly run our personal and professional lives through online platforms and SaaS products, yet our digital identities are still largely protected by outdated, vulnerable methods like usernames, passwords, and easily phishable MFA methods. As long as these remain the primary means of access, attackers will continue to exploit them at scale with infostealer malware and phishing. This highlights the urgent need for adoption of stronger, phishing-resistant authentication methods, continuous identity monitoring, and a shift toward identity-centric security models. "It also reinforces the need for organizations to adopt an identity-centric security posture and monitor for malicious activity even when logins appear legitimate. In today's SaaS driven environments, users and systems authenticate from anywhere, often using federated identity providers like Apple, Google, and Meta. This makes identity a primary control point for security." Both incidents reveal the critical need for vigilance and adaptation in security practices, as threat actors continue to exploit outdated habits and overlooked vulnerabilities with increasing effectiveness and reach.
Techday NZ
an hour ago
- Techday NZ
Most high-traffic email domains still vulnerable to phishing
New research from EasyDMARC has found that 92% of the world's top 1.8 million email domains lack adequate protection against phishing attacks. The EasyDMARC 2025 DMARC Adoption Report has revealed that only 7.7% of these domains are fully protected using the strictest DMARC (Domain-based Message Authentication, Reporting, and Conformance) policy, known as 'p=reject'. This policy is designed to actively block malicious emails from being delivered to inboxes. DMARC is an email authentication protocol that builds on existing standards such as SPF and DKIM, allowing domain owners to specify how they want mail servers to handle emails that fail authentication checks. The protocol also enables domain owners to receive reports on emails sent under their domain name, providing vital records of authentication attempts and potential abuse. EasyDMARC's analysis demonstrates that although there has been a noticeable increase in DMARC adoption since 2023 — largely due to regulatory initiatives and mandates from major providers including Google, Yahoo, and Microsoft — most organisations opt for the weakest available configuration, 'p=none'. This setting only monitors for threats, rather than thwarting attacks by blocking illegitimate emails. The report, which reviewed security practices across the most-visited websites globally as well as Fortune 500 and Inc. 5000 companies, shows a continued gap between DMARC adoption and meaningful implementation. More than half (52.2%) of the surveyed domains have not implemented DMARC at any level, leaving them exposed to phishing and spoofing risks. Among domains that do have a DMARC record, most have not configured enforcement policies or reporting mechanisms necessary for full protection. The research also found that over 40% of the domains with a DMARC record did not include any reporting tags. This omission means these organisations have little to no visibility into authentication failures or an understanding of who might be sending emails on their behalf. Gerasim Hovhannisyan, Chief Executive Officer of EasyDMARC, addressed the misconception surrounding DMARC adoption: "There's a growing perception that simply publishing a DMARC record is enough. But adoption without enforcement creates a dangerous illusion of security. In reality, most organisations are leaving the door wide open to attacks targeting customers, partners, or even employees." Mandates have had a measurable effect. In the United States, where regulatory enforcement is strong, the proportion of phishing emails accepted dropped from 68.8% in 2023 to just 14.2% in 2025. Similar progress was noted in the UK and the Czech Republic, countries that also enforce DMARC usage. However, countries without strict requirements, such as the Netherlands and Qatar, showed minimal improvement in reducing phishing acceptance rates. Recent high-profile cyber attacks, including those targeting retailers such as M&S and Co-op, serve as a backdrop for the report's release. In these incidents, attackers exploited weaknesses in email security through social engineering, costing affected businesses hundreds of thousands in losses. According to EasyDMARC, the increasing sophistication of phishing, partly driven by the use of AI, magnifies the risks for organisations that are inadequately protected. Hovhannisyan further commented: "Misconfigurations, missing reporting, and passive DMARC policies are like installing a security system without ever turning it on. Phishing remains one of the oldest and most effective forms of cyberattack, and without proper enforcement, organisations are effectively handing attackers the keys to their business. As threats grow more sophisticated and compliance pressures mount, stopping halfway with DMARC enforcement is no longer an option." The report methodology combined public DNS data with proprietary data collected through EasyDMARC's platform. It involved the review of aggregate DMARC reports from major mailbox providers and included a survey of 980 IT professionals across the United States, United Kingdom, Canada, and the Netherlands. This allowed for insights into regional differences in phishing trends, adoption challenges, and the varying influence of regulatory mandates. The research concludes that while DMARC adoption has increased, genuine protection against phishing relies on both enforcement and visibility — elements still missing for the vast majority of high-traffic domains worldwide.
1News
a day ago
- 1News
Two Kiwi students shine in Apple's global coding competition
Two 16-year-olds from New Zealand have been named among Apple's Swift coding competition winners for 2025. Alex Liang from Westlake Boys High School and Ben Lawrence from Kaiapoi High School were named among the 350 winning submissions from the tech giant's global competition associated with the Worldwide Developers Conference. Applicants span the globe, representing 38 countries and regions, and incorporating a wide range of tools and technologies. Liang's entry called Make A Wish follows his success last year as the only winner from New Zealand, this time using maths to predict a meteor path, track it in the sky and capture pictures of meteor showers from a phone, rather than a meteor camera. He said he vividly remembered standing outside in May 2021, holding his mother's phone up for three hours and being "very happy and proud of myself" when he finally captured a meteor on camera — but it was time consuming. ADVERTISEMENT "My app is focused on meteors because many people actually try to see meteors and then try to take a picture of it. Right now, the cheapest way to capture a meteor without having to endure it is to use a meteor camera." Alex Liang demonstrates his app to a fellow student. (Source: Supplied) For Liang, it's the second year in a row that he has been named as a winner and he said there had been a "very clear rise" in the number and calibre of applicants this year, and in the use of Artificial Intelligence. "And so this year, unlike last year, I mentioned little planets, which was no AI at all. But this year I did actually implement AI/machine learning in the form of object detection." Liang said he was "not surprised at all" that Apple had decided to allow the use of AI in the competition this year but all usage had to be disclosed. "In fact, I was expecting it. Without AI, many of the things from the app I made would not be possible. "I use it to bump up my efficiency sometimes because I do actually use AI to debug and stuff and then sometimes to create new features or learn new frameworks. But using it does not mean you have to rely entirely on it. ADVERTISEMENT "You have to understand your code top to bottom, every single line, not just saying 'hey GPT do something for me, just write me an app that's doing this'. And AI is not able to create things like that just by saying one word." Liang said planning was already underway for next year's entry, and Make A Wish was being reviewed by Apple to be added to the App Store. "Words cannot describe the experience. I felt like I stepped up to the whole next level of not just astronomy, but innovation. It is something I'm very profound about, something I'm very happy about." Helping money make 'Good Cents' Ben Lawrence pictured using his app, Good Cents. (Source: Supplied) 16-year-old Ben Lawrence from Kaiapoi High School, Christchurch, told 1News he "didn't expect anything" after entering his submission, Good Cents. The app simulates real-world financial scenarios in which users get a job, spend and save money and navigate complex financial curveballs. ADVERTISEMENT 'You do a quiz on some financial questions like 'What is a good way to spend money?' 'How do you save?' 'What's a budget used for?'. 'And based on that, you'll be awarded points, you'll get promotions and the player can also complete lessons that will teach you certain elements and aspects and then I'll quiz you on it to make sure you've actually read through it." Throughout the game, Lawrence gave examples of "random events" players could encounter designed to test whether they will spend money or not. "Oh, the new iPhone came out. You know you already have one, but you want the new one. Do you want to do it, or should you save your money? That kind of thing." Ben Lawrence's app, 'Good Cents'. (Source: Supplied) "Or if your savings are low but you have tonnes of money in your spending [the app] might say, 'hey do you want to put some money into your savings, get some interest on it?'." Lawrence said learning money skills in class could be "pretty boring" and hoped his app could be a way to make learning finances more fun. ADVERTISEMENT "Just making learning more fun and then also helping people with skills better pretty darn important and going into adulthood." The app took him three months to develop, and he hoped to launch it on the App Store soon. 'But I'm working on kind of upgrading it, almost making it so it's more of a platform so schools can sign up to it, license it, whatever and you can have classrooms and teachers can assign work to students and certain aspects of it. 'Phones are banned in schools so that's a huge problem, but I'm working on making it so that they can do it through a website now as well.' An idea that could go 'global' Denis Vida, Professor of Physics and Astronomy at Western University in Ontario Canada, runs a project called the Global Meteor Network (GMN), which has over 1400 meteor cameras globally across 42 countries. Liang said he was collaborating the organisation with the goal of sharing his app with the global astronomy community. ADVERTISEMENT Adjunct Research Professor at the University of Western Ontario. (Source: Supplied) "Essentially everyone can follow a simple set of instructions and buy very reasonably low-cost hardware and install a meteor camera, then install our open-source software and contribute to the project," Vida said. He said "you don't really need to know much about space or science" to do so. "We have a lot of participants in New Zealand who are farmers with no previous context or interest in astronomy, but when they heard about potential meteor fall in their area they got in touch to get a camera installed." He said Liang reached out to him as one of New Zealand's "strong, well-organised" group of space enthusiasts, and was excited to collaborate with him on a project to solve a common problem. Vida said the main issue they faced was that people may be told when a meteor event could happen in their area, but won't know where exactly to look or which way to point their phone in the sky. "Most of the time when people do it, they hold their phone up, they wait for the right time, and it turns out they were looking the completely wrong direction. Or they'll swing the phone and then the only thing they have in the frame is just the fireball with no other reference points. ADVERTISEMENT "The problem is if we want to make measurements in that, we can't use it. There needs to be stars, it needs to be static or have some kind of reference points." He said developing an app to solve this problem sounded like the perfect project for programming pro Liang. "So the idea was let's build an app or some service that people can install where every time something like that happens within a certain radius of you, you'll get a notification and a set of instructions of how to start calibration and the sorts of images to take." Vida said now that Liang had created the app, the next step was getting it installed on a lot of phones, to get other meteor agencies on board. "Once we know that, you know things are going to happen and then people are going to install it, they're going to take pictures with the app and then once we show results that's where the app is going to get more established or where people are going to get more recognised." He said with some more rigorous testing and development of some features, Liang's app would have the potential to have an "oversized impact" globally. "Impacts of little asteroids happen all over the world in a random way. So something's on the App Store and anyone can download it, that's literally [going to] go global." ADVERTISEMENT "These challenges are a great way to find talent. You have pretty small investment and you set some sort of a goal, and you find talented people who are inspired by it," he said.
