Ethical Risks of Deepfake Technology

AI PR Nightmares Part  3- Deep Fakes Will  Strike Deeper (start planning now): Cyber tools that clone voices and faces arent social media scroll novelties, they’re now mainstream weapons causing millions or billions in financial and reputational harm. If you haven’t scenario‑planned for them yet, you have some work to do right Video, audio, and documents so convincing they could collapse reputations and finances overnight. This isn’t distant Sci‑Fi or fear mongering: Over 40% of financial firms reported deep‑fake threat incidents in 2024 and it escalated 2,137% in just three years. 😱 ⚠️ Real-world fraud: The CFO deep‑fake heist: In early 2024, a British engineering firm (Arup) fell victim to a video‑call deepfake featuring their CFO. Scammers walked an employee through 15 urgent transactions, ultimately siphoning off over $25 million. This wasn’t social media fakery, it was a brazen boardroom attack, executed in real time, with Cold War KGB‑level human believability. 🎭 What synthetic mischief will look like tomorrow: 😱 Imagine a deep‑fake video appearing of a Fortune 500 CEO allegedly accepting a bribe, or footage showing them in inappropriate behavior. 😱 And then within minutes it’s gone viral on social and in the mainstream press, before the real person or company one can even issue a statement. The 2025 version of Twain’s “a lie can travel halfway around the world before the truth puts on its shoes”, except a 1000X faster. At that point, the reputational damage is done even if the clip is later revealed as AI‑generated. 🛡️ What companies must be doing now: Audience Action: Internal (Staff): - Run mandatory deepfake awareness training. - Tell teams: “Yes, you might get a video call from your boss, but if it’s not scheduled, don’t act, and verify via text, email or call. Investors & Regulators: - Include a standard disclaimer in all earnings and executive communications: - “Any video/audio statements are verified via [secure portal/email confirmation]. If you didn’t receive a confirmation, assume it’s fake.” Customers & Partners: - Publish your deep‑fake response plan publicly; kind of like a vulnerability disclosure for your reputation. - Say: “We will never announce layoffs or major program changes via a single email/video.” Media & Public: - Pre‑train spokespeople to respond rapidly: - “That video is fraudulent. We’re initiating forensic authentication and investigating now.” Digital Defense: - Invest in deep‑fake detection tools. Sign monitoring agreements with platforms and regulators. Track your senior execs’ likenesses online. 👇 Has your company run deep‑fake drills? Or do you have a near‑miss story to share? Let’s all collaborate on AI crisis readiness.

Seeing the Ghibli memes flood my feed with people’s teens and young children has made me feel really uncomfortable. As I don’t see this reported often, I hope it’s ok to write about it this once— Currently, people are generating so much Gen AI CP that it’s become increasingly difficult for law enforcement to find & rescue real human child victims. While tools could cause harm before, gen AI now produces CSAM in tens of thousands of images by just one person—at unprecedented speed & scale. Law enforcement sifting through 100M pieces of CSAM must now determine: 1. If content features a real child 2. If a child’s identity was concealed with AI 3. If depiction of act is real When you post daily photos of your child on social media, these photos could be used to generate non-consensual explicit images & videos used for: - Grooming - Blackmail - Financial extortion - Bullying In Thorn's new report, it said that gen AI is now increasingly used by minors under 18 to create harmful deepfakes of their peers. The normalization of these tools used 'for fun' to manipulate their likeness have increased its usage—where "1 in 8 minors say that they knows someone who has created deepfake nudes of others", adding another layer of danger spreading through schools & local communities. In 2023, National Center for Missing & Exploited Children received more than 100 million pieces of suspected CSAM. While AIG-CSAM is a small fraction—it strains law enforcement’s ability, & impedes the help & rescue for real victims. The proposed solution so far is more AI. But no algorithm can remove the trauma once a victim experiences it. Better identification won’t change that gen AI enables new, irreversible harms at unprecedented scale. Every time you help an AI model to viral—it encourages companies to deploy faster, more powerful ones. Without addressing harms & risks this technology carries, we’re supporting the speed up of this harm. Why must we rush to adopt these technologies without considering human costs? What makes generating a Ghibli meme of your child, in exchange for the harm it can do for them, worth it? It’s one thing to say you’re required to use it at work—but to normalize gen AI for manipulating you & your children’s pictures—have you explained to them what it does, the impact it may have on their lives in the future? Do you give them a choice to say no? — Many may not know about artist copyright issues, but gen AI poses risks for everyone. For those wanting to use it for fun, I think at the bare minimum, they need to be informed on the risks & improve safety for themselves & their families. I put this together so anyone who wants to share it with others who may not know about these harms, can do so and let them be aware. *Full post & links on my blog (TW beware): https://lnkd.in/gSMQCHpn

Artificial Intelligence (AI) tools are being used by cybercriminals to trick victims. How effective are AI cloned voices when used for fraud? AI voice cloning can replicate human speech with astounding accuracy, revolutionizing industries like entertainment, accessibility, and customer service. I took some time to experiment with an AI voice cloning tool and was impressed with what these tools can do. Using a small voice sample that could be obtained from social media or a spam call, anyone's voice can be cloned and used to say anything. The cloning even includes filler pauses and "umms." This technology powers lifelike virtual assistants and engaging audiobooks, but carries high potential for abuse. Deepfake voice recordings, impersonation, and disinformation campaigns are real concerns. A person's voice can no longer be trusted. A criminal may use a voice that sounds almost identical to your friends or family members. For $1 I had the ability to clone any voice and use it to speak whatever I wanted. I tested with my own voice, and it was eerily realistic. In the age of AI voice cloning software that can enable malicious activities, be vigilant. When answering calls from unfamiliar numbers, allow the caller to speak first. Anything you say could become an audio sample to impersonate you. Consider using a prearranged code word with friends and family as an extra layer of verification. The FTC recommends alternative verification methods, like calling the person on a known number or reaching out to mutual contacts if you suspect a scam. #AI #VoiceCloning #Cybersecurity #Deepfakes #SecurityAwareness