Deepfake Technology Issues

4 billion people go to polls for elections in 2024, and the world is bracing itself for the impact of deep fakes. On January 10 in the US, we saw the introduction of the No Artificial Intelligence Fake Replicas And Unauthorised Duplications (No AI FRAUD) Act, addressing the challenges posed by AI deep fakes in the realm of personal identity and IP. It aims to establish a comprehensive federal framework to protect Americans’ rights to their likeness and voice against AI-generated forgeries. This legislative effort is a response to the misuse of AI in cloning and impersonating individuals, affecting not just celebrities but citizens at large. High-profile cases have shown AI-generated content misrepresenting artists and celebrities, while ordinary individuals have also been victimised by nonconsensual AI-generated content. The No AI FRAUD Act's key features include: - Reaffirming the protection of individual likeness and voice. - Enabling legal action against non-consensual creation and distribution of AI forgeries. - Balancing individual rights with First Amendment protections. It aims to address the shortcomings of state-level laws by setting a federal standard, crucial in an era where digital content transcends state borders. In contrast, the EU's AI Act presents a broader regulatory framework on deep fakes, focusing on the transparency and ethical use of AI. Article 3 (44d) of the EU's AI Act defines 'deep fakes' as manipulated or synthetic content featuring individuals in scenarios they did not actually partake in, created using AI techniques. Article 52(3) mandates disclosure of AI-generated content, requiring users to clearly indicate that the content is artificially generated or manipulated and, where possible, identify the content's creator. The EU's AI Act, however, provides exceptions under Article 52(3a). These exceptions apply when AI-generated content is used under legal authorisation, for freedom of expression, or in the arts and sciences. This clause ensures that the use of AI for creative, satirical, artistic, or fictional works is not unduly restricted, balancing regulatory needs with creative freedom. Comparing the two, the No AI FRAUD Act is more targeted, focusing on the protection of individual rights to likeness and voice, while the EU's AI Act encompasses a broader approach to AI regulation, emphasising transparency and ethical use. The No AI FRAUD Act does not explicitly require disclosure of AI-generated content, highlighting its focus on individual rights over broader AI governance issues. Both pieces of legislation reflect a growing recognition of the challenges AI poses to personal identity and intellectual property. However, their approaches differ in scope and emphasis—while the No AI FRAUD Act centers on safeguarding personal rights in the U.S., the EU's AI Act seeks to establish a regulatory framework for ethical AI use, including transparency and disclosure requirements, across its member states.

When a patient hears from a “doctor,” they shouldn’t have to wonder if it’s real. AI deepfakes are already blurring that line - impersonating physicians, promoting unproven treatments, and putting patients at risk. When a physician’s identity is hijacked, it doesn’t just harm one clinician. It undermines the credibility of care itself. That’s why the AMA is calling for clear, enforceable protections against AI-driven impersonation. We’ve outlined a framework grounded in a simple idea: a physician’s identity is not a public utility. It’s a protected right. What does that mean in practice? • No use of a physician’s name, likeness, or voice without explicit, informed, and revocable consent. • Clear labeling and transparency for any AI-generated clinical content. • Shared accountability across platforms, vendors, and institutions. • Real enforcement mechanisms to stop impersonation and protect patients. This isn’t simply about stopping bad actors. It’s about defining the rules of trust in a digital health system. If identity can be manufactured today, what anchors trust in health care tomorrow? #AI #DigitalHealth #PatientSafety #Deepfakes

The U.S. Copyright Office this month released Part 1 of its report on the legal and policy issues related to copyright and artificial intelligence, focusing on digital replicas. For those who don't know, digital replicas are video, image, or audio recordings digitally created or manipulated to realistically but falsely depict an individual. An example includes the April 2023 song "Heart on My Sleeve," featuring AI-generated voices of Drake and The Weeknd. Here is a topline breakdown of what the Copyright Office has shared: 1️⃣ Existing Legal Frameworks and the Need for Federal Legislation: Current state and federal laws are inadequate in addressing digital replicas, revealing significant gaps and inconsistencies. Privacy and publicity rights are insufficient to tackle AI-generated content. Comprehensive federal legislation is urgently needed to protect individuals from substantial harm. 2️⃣ Scope and Protection of New Law: Proposed legislation should target digital replicas indistinguishable from authentic depictions, offering more precise protection than existing 'name, image, and likeness' laws. This law should apply to all individuals, not just celebrities, and extend at least for a lifetime with possible postmortem extensions. 3️⃣ Infringing Acts and Secondary Liability: Liability should focus on the distribution or availability of unauthorized digital replicas, covering both commercial and non-commercial uses. Traditional tort principles of secondary liability should apply, with a safe harbor mechanism for online service providers to remove infringing content upon notice. This approach balances accountability and practicality for intermediaries. 4️⃣ Licensing, Assignment, and Free Speech Concerns: Individuals should have the right to license their digital replica rights but not permanently assign them to others. Additional safeguards are needed to protect minors from exploitation. The legislation should address free speech concerns by balancing protection of individual rights with the need to avoid overly broad restrictions on expression. 5️⃣ Effective Remedies: The report calls for strong legal remedies, including court orders to stop infringement (injunctive relief) and monetary compensation. It emphasizes the importance of statutory damages and covering legal fees to make the law accessible to everyone. In severe cases, criminal penalties are suggested to act as a strong deterrent against violations, ensuring the law's effectiveness and fairness. Check out the full report below ⤵

As AI tools advance rapidly, it's important for employers to understand where the ethical and legal boundaries lie. The EU AI Act has taken a firm stance: AI systems that infer personality or emotions from biometric data — including face-based personality prediction — are prohibited or classified as high-risk. The legislation recognises the profound risks these tools pose to fairness, discrimination, privacy, and human dignity. In Australia, no equivalent protections currently exist. This means technologies that would be unlawful in Europe could still enter the Australian recruitment market — without the guardrails needed to prevent discrimination or algorithmic bias. As employers explore AI for hiring, screening, or talent management, now is the time to stay alert: —Be cautious of AI tools claiming to “predict personality” or “assess fit” from images or videos. —Demand transparency, validation evidence and bias testing from vendors. —Ensure any AI used in HR aligns with ethical standards — even if legislation lags behind. Until stronger regulation arrives in Australia, the responsibility rests with employers to safeguard their people and their processes from high-risk AI. Join the growing community of multidisciplinary leaders for inclusive and ethical AI at ada.ai.

📡 AI SIGNAL -->𝐃𝐞𝐧𝐦𝐚𝐫𝐤 𝐌𝐚𝐤𝐞𝐬 𝐘𝐨𝐮 𝐈𝐏 — 𝐅𝐚𝐜𝐞, 𝐕𝐨𝐢𝐜𝐞, 𝐚𝐧𝐝 𝐋𝐢𝐤𝐞𝐧𝐞𝐬𝐬 𝐀𝐫𝐞 𝐍𝐨𝐰 𝐏𝐫𝐨𝐭𝐞𝐜𝐭𝐞𝐝 𝐁𝐲 𝐂𝐨𝐩𝐲𝐫𝐢𝐠𝐡𝐭. Denmark just passed a historic law giving citizens copyright over their own biometric identity—including face, voice, and other personal characteristics. This marks a paradigm shift: 🔁 Identity is no longer just data—it's intellectual property. 𝐖𝐇𝐘 𝐓𝐇𝐈𝐒 𝐌𝐀𝐓𝐓𝐄𝐑𝐒 Until now, deepfakes and AI-generated impersonations lived in a gray zone. It was hard to take them down unless fraud or harm could be proven. Now? 🎯 It’s illegal to use someone’s likeness or voice without consent—even if no money changes hands. 🎯 That gives individuals and companies a clear legal route to report, remove, and penalize deepfake content. 𝐖𝐇𝐎 𝐁𝐄𝐍𝐄𝐅𝐈𝐓𝐒 — 𝐀𝐍𝐃 𝐇𝐎𝐖 👵 Seniors: Voice scam victims can now pursue impersonators through copyright law, not just vague fraud claims. 👶 Children: AI-generated imitations of kids without consent become violations—potentially actionable by parents or guardians. 🏦 Banks & businesses: Deepfake identity abuse (like spoofed customers or execs) becomes a compliance and IP issue—not just a technical risk. This doesn’t magically stop scams—but it gives victims and platforms legal power to act quickly and decisively. 𝐋𝐎𝐍𝐆-𝐓𝐄𝐑𝐌 𝐈𝐌𝐏𝐀𝐂𝐓 ✅ Identity becomes programmable, ownable, and enforceable ✅ AI systems will need consent and provenance layers built in ✅ Could accelerate the adoption of decentralized identity (DID) systems and agent-based consent management 𝐁𝐎𝐓𝐓𝐎𝐌 𝐋𝐈𝐍𝐄 This law doesn’t stop deepfakes with code—it stops them with consequences. It moves synthetic identity abuse from a technical curiosity to a compliance violation. And it sends a clear signal: 𝐘𝐨𝐮 𝐚𝐫𝐞 𝐲𝐨𝐮𝐫 𝐨𝐰𝐧 𝐈𝐏.

Do California’s legislative efforts to regulate AI reflect a growing concern for digital ethics, personal rights, and democratic integrity, or are they legislative overreach that will stifle innovation? There are five key bills ready to be signed into law that address issues ranging from the use of digital replicas in contracts and posthumous rights of deceased personalities to the transparency and safety of AI platforms. I’ve read them. You should, too. The links and short descriptions are below. AB 2602: Contracts: digital replicas – This bill mandates that contracts for personal or professional services involving digital replicas must clearly specify the intended uses of the replica. It also requires that individuals involved have access to legal counsel or labor union representation during contract negotiations in order to protect performers’ rights in the digital age. SB 1047: Safe and Secure Innovation for Frontier Artificial Intelligence Models Act – This bill establishes safety regulations for “covered AI models,” defined by computational power and training costs. Developers of these models must implement safety measures, conduct regular audits, and report significant incidents to the California Department of Technology. AB 2013: Artificial intelligence: transparency – This bill requires businesses that use generative AI systems to disclose any use of copyrighted materials in the training data. It also mandates that clear information about the AI system’s capabilities and limitations be provided to users. AB 1836: Deceased personalities: digital replicas – This bill prohibits the use of digital replicas of deceased personalities in audiovisual works without prior consent from their estate. It extends protections to ensure that digital replicas are not used posthumously without authorization, addressing concerns of exploitation. AB 2655: Defending Democracy from Deepfake Deception Act of 2024 – This bill requires clear disclosure of AI-generated content in political advertisements and campaign materials. It also prohibits the distribution of deceptive audio or visual media of candidates that could mislead voters, aiming to protect electoral integrity.

The NO FAKES Act, which faces a key Senate Judiciary Committee vote, addresses one of the most significant gaps in AI law: ownership and control of human identity. Copyright law protects creative works. Trademark law protects brands. But neither was built to address AI-generated replicas of a person's voice or likeness. The NO FAKES Act would establish a federal framework governing unauthorized digital replicas, creating legal rights and remedies where existing laws remain fragmented across states. In doing so, it shifts the conversation beyond AI training data and copyright infringement to a more fundamental question: Who owns your digital identity? While the bill fills an important legal gap, it will likely face debate over how to balance identity protection with free expression, parody, journalism, and innovation. As with many AI laws, the challenge is not defining the problem, it is drawing the line in the right place. As AI-generated content becomes increasingly indistinguishable from reality, legal protections for voice, likeness, and identity may become as important as traditional intellectual property rights. MBW coverage: https://lnkd.in/e8Ax8XKd

Today the UK government published a report on AI and copyright which sets out a policy agenda for the rest of this year, including a summer consultation on digital replicas and a taskforce on labelling AI-generated content, with an interim report due in the autumn. Digital replicas sit right at the intersection of creativity and identity. Used with permission, they can unlock new formats and create new business models as I've argued previously: https://lnkd.in/ey3Xg6S3 The government is right to frame this issue around real-world harm, rather than abstract theories about the underlying technology. The same is true for labelling. Consumers browsing a social newsfeed might benefit from knowing when content is AI generated, and disclosure can help reduce the impact of disinformation and harmful deepfakes; that premise however doesn’t necessarily hold true for businesses using generative media in an enterprise context. Views are divided on this topic but I believe labelling is not a sticker that you slap on everything to magically solve all problems. If we want to protect consumers from harm, an approach to labelling has to be practical, interoperable, proportionate and most importantly paired with a renewed focus on digital literacy. So what we should aim for is best practice that scales across the ecosystem while keeping in mind important differences between consumer and enterprise software or between model providers and application-layer companies. Policies should focus on the contexts where provenance really matters, like consumers being exposed political content, public safety, financial fraud and non-consensual impersonation. They should also avoid turning every legitimate creative workflow into a paperwork exercise, because the people who will struggle most with heavy process are UK AI startups, most of which are application-layer companies developing enterprise software, not AI models or social media apps. With the above context in mind, here's a direction of travel the UK can follow: 1. Protect real people from unauthorized replication and impersonation. 2. Make transparency usable in the places it actually reduces harm. 3. Keep the UK a place where responsible AI companies can build, test and grow, without spending their best engineers reinventing the compliance wheel. If the goal is to make this country the best place to create and the best place to build, an evidence-led approach is how we get there.

U.S. Copyright Office: Copyright and #AI - #Digital #Replicas This first Part of the Copyright Office’s Report on copyright and AI addresses the topic of digital replicas. From AI-generated musical performances to robocall impersonations of political candidates to images in pornographic videos, an era of sophisticated digital replicas has arrived. Although technologies have long been available to produce fake images or recordings, generative AI technology’s ability to do so easily, quickly, and with uncanny verisimilitude has drawn the attention and concern of creators, legislators, and the general public. Section I summarizes the context and history of the Office’s study of the digital replicas issue. Section II.A outlines the main existing legal frameworks: state rights of privacy and publicity, including recent legislation specifically targeting digital replicas, and at the federal level, the Copyright Act, the Federal Trade Commission Act, the Communications Act, and the Lanham Act. Section II.B explains why existing laws do not provide sufficient legal redress for those harmed by unauthorized digital replicas and propose the adoption of a new federal law. Section III discusses protection against AI outputs that deliberately imitate an artist’s style.

A new bill was just introduced in the house that addresses #generativeai and #deepfakes from the standpoint of #rightofpublicity in light of the rash of celebrity impersonations, revenge porn/nonconsensual intimate images, and, presumably, the upcoming election. It's essentially a proposed federal publicity right but limited to the context of digital deepfakes, and is called the No AI FRAUD Act (No Artificial Intelligence Fake Replicas and Unauthorized Duplications Act). I want to know whose job it is to name these bills. The acronyms folks on the Hill come up with are just bonkers. But I digress. High points (#tldr): 1. Would apply to any person, living or dead. 2. Includes protection for unauthorized use of name, voice, and likeness, with likeness broadly defined to also include "face, likeness, or other distinguishing characteristic". 3. Rights are defined as #intellectualproperty rights as opposed to privacy rights (there is an ongoing debate as to where publicity rights fall, this bill plants the flag in the property camp). 4. Rights are descendible and transferrable. 5. Authorized use requires the subject to be 18 and be represented by counsel (an interesting twist) or a collective bargaining agreement. 6. Includes both direct and contributory liability provisions. 7. Damages include the greater of actual damages or $50,000 per violation, plus profits of the violator above actual damages (similar to #copyright actual damages) for digital cloning and $5,000 per violation or actual damages (plus profits) for digital voice replicas. 8. Punitive damages and attorney's fees are available. 9. Disclaimers don't get you off the hook. 10. There are #firstamendment carve-outs that consider commerciality, expressive purpose, and competition/adverse market effects. 11. Limitations on liability where harm is "negligible" (interesting side note - definition of harm includes emotional harm) and where uses are transformative or constitute commentary on matters of public concern. 12. Various types of intimate or sexual images will lead to per se harm. 13. Four-year statute of limitations. It is an ambitious bill and certainly includes provisions I can see being challenged. Who knows if this will go anywhere, but #congress at least has important #ai issues on its radar. #golfclap chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://lnkd.in/eDS23Q_Z