Latest news with #AIcontent
CBC
5 days ago
- CBC
UW researchers figured out how to bypass anti-deepfake markers on AI images
University of Waterloo researchers have built a tool that can quickly remove watermarks identifying content as artificially generated — and they say it proves that global efforts to combat deepfakes are most likely on the wrong track. Academia and industry have focused on watermarking as the best way to fight deepfakes and "basically abandoned all other approaches," said Andre Kassis, a PhD candidate in computer science who led the research. At a White House event in 2023, the leading AI companies — including OpenAI, Meta, Google and Amazon — pledged to implement mechanisms such as watermarking to clearly identify AI-generated content. AI companies' systems embed a watermark, which is a hidden signature or pattern that isn't visible to a person but can be identified by another system, Kassis explained. He said the research shows the use of watermarks is most likely not a viable shield against the hazards posed by AI content. "It tells us that the danger of deepfakes is something that we don't even have the tools to start tackling at this point," he said. The tool developed at the University of Waterloo, called UnMarker, follows other academic research on removing watermarks. That includes work at the University of Maryland, a collaboration between researchers at the University of California and Carnegie Mellon, and work at ETH Zurich. Kassis said his research goes further than earlier efforts and is the "first to expose a systemic vulnerability that undermines the very premise of watermarking as a defence against deepfakes." In a follow-up email statement, he said that "what sets UnMarker apart is that it requires no knowledge of the watermarking algorithm, no access to internal parameters, and no interaction with the detector at all." When tested, the tool worked more than 50 per cent of the time on different AI models, a university press release said. AI systems can be misused to create deepfakes, spread misinformation and perpetrate scams — creating a need for a reliable way to identify content as AI-generated, Kassis said. WATCH | How to spot a deepfake: How to spot a deepfake 4 months ago AI-generated videos have sprung up all over social media. Recently, a scam has been using deepfakes of CBC's David Cochrane and Prime Minister Justin Trudeau. A deepfake of U.S. President Donald Trump and Elon Musk made headlines last week that involved feet. Here are some ways to detect them. After AI tools became too advanced for AI detectors to work well, attention turned to watermarking. The idea is that if we cannot "post facto understand or detect what's real and what's not," it's possible to inject "some kind of hidden signature or some kind of hidden pattern" earlier on, when the content is created, Kassis said. The European Union's AI Act requires providers of systems that put out large quantities of synthetic content to implement techniques and methods to make AI-generated or manipulated content identifiable, such as watermarks. In Canada, a voluntary code of conduct launched by the federal government in 2023 requires those behind AI systems to develop and implement "a reliable and freely available method to detect content generated by the system, with a near term focus on audio-visual content (e.g., watermarking)." Watermark removed 'within 2 minutes max' Kassis said UnMarker can remove watermarks without knowing anything about the system that generated it, or anything about the watermark itself. "We can just apply this tool and within two minutes max, it will output an image that is visually identical to the watermark image" which can then be distributed, he said. "It kind of is ironic that there's billions that are being poured into this technology and then, just with two buttons that you press, you can just get an image that is watermark-free." Kassis said that while the major AI players are racing to implement watermarking technology, more effort should be put into finding alternative solutions. Watermarks have "been declared as the de facto standard for future defence against these systems," he said.
CTV News
5 days ago
- Business
- CTV News
Canadian researchers create tool to remove anti-deepfake watermarks from AI content
OpenAI CEO Sam Altman participates in a panel discussion during the annual meeting of the World Economic Forum in Davos, Switzerland, Thursday, Jan. 18, 2024. OpenAI was one of the major tech firms that promised to pursue watermarking technology. (AP Photo/Markus Schreiber) OTTAWA — University of Waterloo researchers have built a tool that can quickly remove watermarks identifying content as artificially generated — and they say it proves that global efforts to combat deepfakes are most likely on the wrong track. Academia and industry have focused on watermarking as the best way to fight deepfakes and 'basically abandoned all other approaches,' said Andre Kassis, a PhD candidate in computer science who led the research. At a White House event in 2023, the leading AI companies — including OpenAI, Meta, Google and Amazon — pledged to implement mechanisms such as watermarking to clearly identify AI-generated content. AI companies' systems embed a watermark, which is a hidden signature or pattern that isn't visible to a person but can be identified by another system, Kassis explained. He said the research shows the use of watermarks is most likely not a viable shield against the hazards posed by AI content. 'It tells us that the danger of deepfakes is something that we don't even have the tools to start tackling at this point,' he said. The tool developed at the University of Waterloo, called UnMarker, follows other academic research on removing watermarks. That includes work at the University of Maryland, a collaboration between researchers at the University of California and Carnegie Mellon, and work at ETH Zürich. Kassis said his research goes further than earlier efforts and is the 'first to expose a systemic vulnerability that undermines the very premise of watermarking as a defence against deepfakes.' In a follow-up email statement, he said that 'what sets UnMarker apart is that it requires no knowledge of the watermarking algorithm, no access to internal parameters, and no interaction with the detector at all.' When tested, the tool worked more than 50 per cent of the time on different AI models, a university press release said. AI systems can be misused to create deepfakes, spread misinformation and perpetrate scams — creating a need for a reliable way to identify content as AI-generated, Kassis said. After AI tools became too advanced for AI detectors to work well, attention turned to watermarking. The idea is that if we cannot 'post facto understand or detect what's real and what's not,' it's possible to inject 'some kind of hidden signature or some kind of hidden pattern' earlier on, when the content is created, Kassis said. The European Union's AI Act requires providers of systems that put out large quantities of synthetic content to implement techniques and methods to make AI-generated or manipulated content identifiable, such as watermarks. In Canada, a voluntary code of conduct launched by the federal government in 2023 requires those behind AI systems to develop and implement 'a reliable and freely available method to detect content generated by the system, with a near-term focus on audio-visual content (e.g., watermarking).' Kassis said UnMarker can remove watermarks without knowing anything about the system that generated it, or anything about the watermark itself. 'We can just apply this tool and within two minutes max, it will output an image that is visually identical to the watermark image' which can then be distributed, he said. 'It kind of is ironic that there's billions that are being poured into this technology and then, just with two buttons that you press, you can just get an image that is watermark-free.' Kassis said that while the major AI players are racing to implement watermarking technology, more effort should be put into finding alternative solutions. Watermarks have 'been declared as the de facto standard for future defence against these systems,' he said. 'I guess it's a call for everyone to take a step back and then try to think about this problem again.' This report by The Canadian Press was first published July 23, 2025. Anja Karadeglija, The Canadian Press
Yahoo
5 days ago
- Science
- Yahoo
Canadian researchers create tool to remove anti-deepfake watermarks from AI content
OTTAWA — University of Waterloo researchers have built a tool that can quickly remove watermarks identifying content as artificially generated — and they say it proves that global efforts to combat deepfakes are most likely on the wrong track. Academia and industry have focused on watermarking as the best way to fight deepfakes and "basically abandoned all other approaches," said Andre Kassis, a PhD candidate in computer science who led the research. At a White House event in 2023, the leading AI companies — including OpenAI, Meta, Google and Amazon — pledged to implement mechanisms such as watermarking to clearly identify AI-generated content. AI companies' systems embed a watermark, which is a hidden signature or pattern that isn't visible to a person but can be identified by another system, Kassis explained. He said the research shows the use of watermarks is most likely not a viable shield against the hazards posed by AI content. "It tells us that the danger of deepfakes is something that we don't even have the tools to start tackling at this point," he said. The tool developed at the University of Waterloo, called UnMarker, follows other academic research on removing watermarks. That includes work at the University of Maryland, a collaboration between researchers at the University of California and Carnegie Mellon, and work at ETH Zürich. Kassis said his research goes further than earlier efforts and is the "first to expose a systemic vulnerability that undermines the very premise of watermarking as a defence against deepfakes." In a follow-up email statement, he said that "what sets UnMarker apart is that it requires no knowledge of the watermarking algorithm, no access to internal parameters, and no interaction with the detector at all." When tested, the tool worked more than 50 per cent of the time on different AI models, a university press release said. AI systems can be misused to create deepfakes, spread misinformation and perpetrate scams — creating a need for a reliable way to identify content as AI-generated, Kassis said. After AI tools became too advanced for AI detectors to work well, attention turned to watermarking. The idea is that if we cannot "post facto understand or detect what's real and what's not," it's possible to inject "some kind of hidden signature or some kind of hidden pattern" earlier on, when the content is created, Kassis said. The European Union's AI Act requires providers of systems that put out large quantities of synthetic content to implement techniques and methods to make AI-generated or manipulated content identifiable, such as watermarks. In Canada, a voluntary code of conduct launched by the federal government in 2023 requires those behind AI systems to develop and implement "a reliable and freely available method to detect content generated by the system, with a near-term focus on audio-visual content (e.g., watermarking)." Kassis said UnMarker can remove watermarks without knowing anything about the system that generated it, or anything about the watermark itself. "We can just apply this tool and within two minutes max, it will output an image that is visually identical to the watermark image" which can then be distributed, he said. "It kind of is ironic that there's billions that are being poured into this technology and then, just with two buttons that you press, you can just get an image that is watermark-free." Kassis said that while the major AI players are racing to implement watermarking technology, more effort should be put into finding alternative solutions. Watermarks have "been declared as the de facto standard for future defence against these systems," he said. "I guess it's a call for everyone to take a step back and then try to think about this problem again." This report by The Canadian Press was first published July 23, 2025. Anja Karadeglija, The Canadian Press Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
Yahoo
11-07-2025
- Business
- Yahoo
Meta Shares Updated Reels Ads Guide, Including Creative Tips
This story was originally published on Social Media Today. To receive daily news and insights, subscribe to our free daily Social Media Today newsletter. Looking to add Reels into your digital marketing mix? You should be. Reels are the most engaging content format across Facebook and Instagram, with video now making up over 60% of all time spent in each app. The growth of short-form entertainment has been a seismic shift in the online landscape, and if you want to maximize your opportunities for attention and reach, you should be looking to integrate Reels into your creative flow. And with more options than ever to create content, even AI material that you don't need to film yourself, there's really no excuse not to be posting Reels from your business account. And if you are thinking about your Reels options, Meta's latest Reels guide could give you the inspiration you need to get started. Meta's updated Reels Ads Guide is a short, 6-page overview of Reels promotions best practices, including data on Reels ads performance: There are also creative and presentation tips to ensure you get the most out of the format: As well as narrative pointers based on what works in video posts: There are also tips on working with collaborators, and the benefits of partner promotions within Reels: There's a range of insights here, and while much of it you may have heard or read before, the fact that these best practice tips are being reinforced by Meta itself, and are therefore based on actual performance data, makes this guide, in particular, a more valuable reference point. And as noted, you can also now use AI tools to generate short video clips, like the latest options in Meta's Edits video editing app, while you can even use other platforms' tools, like Google Veo, which enables full video generation from text prompts. There are more options on this front than ever before, and with these additional Reels posting tips, it could end up being a valuable element in your strategy. You can download Meta's Reels Ads Guide here. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
Forbes
28-05-2025
- Business
- Forbes
5 Ways To Make Your ChatGPT-Generated Content Undetectable
5 ways to make your ChatGPT-generated content undetectable People think they know when content is AI-generated. The em dash comes under scrutiny almost daily on LinkedIn, with LinkedIn gurus claiming to have spotted the number one culprit that you're using AI to write. The problem is not em dashes. They have been used in literature for centuries. ChatGPT and other large language models were trained on billions of words from across the internet. Their training data includes em dashes, so the output does too. But em dashes in themselves are not a sign of AI-generated content. The real giveaways run deeper. Here's how to spot them, and there's what to add to your prompts so you don't fall into the traps. Real journalists don't use title case. It's marketing speak that invades landing pages and infects tweets. BBC news uses only sentence case. Title case is thoroughly unsystematic, dubbed 'sensationalist case' by those in the know. AI programs love to use title case and they won't stop unless you tell them to. Most require a comprehensive explanation of what sentence case even means before they'll give it to you. OpenAI CEO Sam Altman knows what's going on. He uses only lower case letters in every tweet. Be less robot and more you. Add this to your prompt: 'For all headings, sections and subsections, use sentence case instead of title case. In sentence case you only capitalise the first letter of a new sentence, and any proper nouns (no common nouns require capitalisation).' AI follows patterns. It creates structures you don't see, robotic rhythms you didn't want, and leads your readers down a predictable path. Break away. You need to mix it up and be free. Follow a proven format, sure. But ad lib within the constraints. Be random, be off-the-wall. Apply perplexity and burstiness. Every now and then, finish a sentence the way no one thought it octopus. Add in words you just made up, let your mind wander, go way off track before you bring it back round to a conclusion. Sometimes, don't even conclude. Add this to your prompt: "Use varied sentence lengths and complexity for dynamic flow. Avoid predictable patterns. Apply perplexity and burstiness. Mix very short sentences with longer ones. Don't follow formulaic structures. Be unpredictable in word choices." In the English language, you only capitalise proper nouns. Nothing else. Not seasons, not topics, even job titles don't need capital letters. Real writers know this, but AI tools don't. Generate a press release or social media post with ChatGPT and you'll spot unnecessary capital letters everywhere. Because it doesn't know the rules. It's making predictions based on its training data, which includes a lot of people getting it wrong. And it's getting worse. Fix it: 'Only capitalise proper nouns. Not job titles, not seasons, not departments or subjects.' When you're talking with friends, you don't hedge your words. You don't add fillers before you get to the sauce. AI does. "Quite possibly," "it could be argued," "somewhat," "rather interesting." Real writers make bold statements and back them up. Robots go in gingerly, waiting for approval. When you're an expert, act like one. Don't apologise for your expertise. Make claims and take stands. Be confident in your knowledge. Don't let AI water you down. Stop qualifying everything you say. Add this to your prompt: "Be direct and assertive. Avoid hedge words like 'quite,' 'rather,' 'somewhat,' 'it could be argued.' Make bold statements without unnecessary qualifiers." "In today's fast-paced digital world" might be the most overused phrase at the start of AI-generated content. Right next to "In conclusion, it's clear that" and "At the end of the day" at the end. Cringe city. Pure robot speak, in proud display in LinkedIn posts, blog articles, and marketing emails. Stop using these terms. Avoid throat-clearing phrases that say nothing. Instead, start your story in the middle of the action. Borrow literary techniques from great filmmakers. Study copywriting formulas that teach you how to hook an audience's attention. Leave them wanting more, not wishing they hadn't bothered. Add this to your prompt: "Skip generic introductions and conclusions. Start with impact, not setup. End with energy, not summary. Avoid phrases like 'in today's world' or 'in conclusion.'" Your ideas deserve better than robotic delivery. These prompt additions turn your AI-generated content from obviously artificial to genuinely engaging. Avoid title case, predictable structures, marketing-jargon capital letters, hedge words and generic openings and endings. Stop worrying about fooling people by using AI. Start producing content you're actually proud of. Content that serves your audience. Content that shares your ethos and strong beliefs. Content that sounds like you. I don't care if your content is AI-generated. I care if it's good. I care that you have prompted compassionately, so you don't waste my time and everyone else's. Access all my best ChatGPT content prompts.
