ContentLens

A single letter changed. A logo close enough to fool a distracted buyer. A product image lifted from a legitimate brand's catalogue. A name that sounds just familiar enough. This is not sophisticated. It does not need to be. Brand identity attacks on e-commerce and social platforms thrive on volume and inattention — not technical sophistication. By the time a brand's legal team notices, thousands of buyers have already been deceived. 𝗪𝗵𝗮𝘁 𝗯𝗿𝗮𝗻𝗱 𝗽𝗿𝗼𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝘀 Brands onboard their assets with ContentLens — logo files, colour palettes, product imagery, packaging typography, trademark registrations. These become the reference layer monitored continuously across social platforms, e-commerce listings, and the open web. 𝗗𝗲𝘁𝗲𝗰𝘁 — ContentLens flags AI-generated content, AI-edited or manipulated visuals, logo and packaging lookalikes, brand name and handle mimics, and unauthorised use of authentic brand assets — wherever they appear online. 𝗖𝗹𝗮𝘀𝘀𝗶𝗳𝘆 — Each detection is assigned an intent. Market Manipulation. Reputation Attack. Promotional or Commercial Gain. Brand Lookalike. Satire or Parody. The classification determines urgency and the action pathway — market manipulation triggers immediate legal escalation; satire may warrant no action at all. 𝗥𝗲𝘃𝗶𝗲𝘄 — ContentLens surfaces findings with full context and evidence. The brand owner — or their legal and communications team — decides what warrants action and what doesn't. 𝗔𝗰𝘁 — Takedown requests to platforms, IP registry escalations, seller account reports, evidence packages structured for litigation. The goal is removal, not negotiation. 𝗣𝗿𝗲𝘀𝗲𝗿𝘃𝗲 — Timestamped, admissible evidence organised for IP litigation or platform dispute resolution. What ContentLens has found, active and publicly listed: 𝗖𝗵𝗼𝗺𝗲𝗰𝗮𝘀𝘁 — one letter off Chromecast, near-identical packaging. 𝗔𝗻𝘆𝗖𝗮𝘀𝘁 — fake product imagery lifted from a legitimate brand. 𝗚𝗨𝗖𝗜𝗜, 𝗚𝘂𝗰𝗰𝗶𝗶, 𝗚𝘂𝗰𝗰𝘆 — three separate variations of GUCCI Flora perfume, each slightly different, all unauthorised. 𝗚𝗦𝟱 — a PlayStation lookalike sold as a gaming console. A night-vision camera lookalike — near-identical to a well-known brand, listed at a fraction of the price. Five violations. Multiple brands. All live. ContentLens is watching. Rohan Sahu | Shivi Mithal #DeepfakeDetection #BrandProtection #IdentityProtection #IPProtection #BrandIntegrity #OnlineFraud #ContentLens

Something using your client's face just went live. Here is what happens next. First, ContentLens catches it. Not because someone searched for it. Because ContentLens was already watching — trained on your client's face, voice, and brand identity from the moment they were onboarded. Reference samples. Precise matching. No keyword required. Second, ContentLens classifies it. Not every hit is a crisis. A meme is not a scam. A parody is not a reputation attack. ContentLens categorises each detection — Promotional, Meme, Reputation Attack, Scam, Counterfeit — before it reaches a human. The reviewer gets context, not just an alert. Third, a human reviews it. This step matters more than most people realise. Wrongful takedowns of satire or legitimate commentary can themselves become headlines. At ContentLens, nothing is actioned without a field officer reviewing and approving first. Speed without judgement is not protection. Fourth — and this is where it gets interesting. If the hit is Promotional — a real product being advertised using an AI-generated likeness of your client — the response is not automatically a takedown notice. It is a licensing conversation. Someone already built an ad campaign around your client's face. The product is real. The commercial intent is real. The only thing missing is your client's consent — and their fee. ContentLens surfaces that moment. What happens next is a business decision, not just a legal one. Fifth, if takedown is the right call, ContentLens initiates it — through platform channels and legal notices, with every piece of evidence stored in a documented chain of custody. For platform appeals, civil suits, or law enforcement escalation, that evidence chain is everything. Detection is just the beginning. The workflow is where outcomes are decided. What would your team do differently if you had this process in place today? Rohan Sahu, Shivi Mithal #DeepfakeDetection #BrandProtection #IdentityProtection #CelebrityProtection #AIThreats #RepresentationRights #ContentLens #BrandIntegrity #OnlineFraud #IPProtection

Your detection tools track keywords. Deepfakes don't use them. The call came on a Monday morning. A manager at a PR and communications agency picked up the phone. A journalist was asking for comment on a video — their client, a well-known public figure, apparently endorsing a financial product. The video had been live since Friday evening. By Monday morning, it had millions of views. The agency had search alerts set up. They had a social listening tool that tracked sentiment. They had a team that checked mentions every morning. None of it caught the video. Because search alerts track keywords, not faces. Social listening tools measure tone, not authenticity. Manual checks cover what you search for — not what you don't know to look for. The video didn't tag the client. It didn't use their name in the caption. It just used their face. And their voice. This is the gap. The threat has evolved. The tools haven't. A deepfake doesn't announce itself. A brand lookalike doesn't send a notification. A voice clone doesn't show up in your mentions feed. One day it is an investment scheme. The next, a gambling app. The day after, a product your client never agreed to sell. They live in places your current stack was never designed to reach — Telegram groups, obscure e-commerce listings, regional platforms, look-alike domains registered overnight. For agencies managing multiple high-profile clients, the bandwidth reality compounds this further. You cannot manually surveil every surface, every platform, every day, for every client. The question is no longer whether your client will be targeted. The question is whether you find out before the journalist calls — or after. At ContentLens, we built our detection on a different foundation entirely — starting with who your client is, not what you happen to search for. What does your current detection stack actually catch — and what does it miss? Rohan Sahu, Shivi Mithal #DeepfakeDetection #BrandProtection #IdentityProtection #CelebrityProtection #AIThreats #RepresentationRights #ContentLens #BrandIntegrity #OnlineFraud #IPProtection

Your face or your brand. Their Profit. Your loss. Ø Last month, actress Rukmini Vasanth woke up to AI-generated images of herself flooding Instagram and X — fabricated, intimate, designed to humiliate. She had never posed for them. Bengaluru Cyber Crime Police filed an FIR against 29 accounts. The damage, however, was already done. This is not just an Indian problem. Ø In 2024, a slick advertisement appeared across social media platforms featuring Taylor Swift — her face, her voice, her unmistakable credibility — announcing she was giving away Le Creuset cookware sets. The brand was real. The giveaway looked legitimate. Thousands engaged. Those who followed the link were charged a hidden monthly fee on their credit cards. Taylor Swift had nothing to do with it. Her identity was simply the most valuable thing the scammers had access to — and it cost them nothing to use. Ø Closer to home, multiple videos of Finance Minister Nirmala Sitharaman began circulating — her face, her voice, her authority — advising ordinary Indians where to invest. Fact-checkers confirmed 99.9% deepfake probability. The linked websites were rated highly dangerous. A man in Mangaluru lost ₹22 lakh trusting one of them. And it is not just people. Ø Search "GUCII Flora" on a major Indian e-commerce platform today. Real bottle, real listing, real sale — built to look and sound exactly like a Gucci fragrance. Gucci sees none of that revenue. Most buyers never notice. Three celebrities. One brand. Four attacks. One common thread: by the time anyone found out, the content had already reached millions. Detection needs to happen before the damage — not after. This is what we are building at ContentLens. Which of these attack types concerns you most — for your brand or your clients? Rohan Sahu | Shivi Mithal #DeepfakeDetection #BrandProtection #IdentityProtection #CelebrityProtection #AIThreats #RepresentationRights #ContentLens #BrandIntegrity #OnlineFraud #IPProtection

Three posts ago, we described a problem most hiring teams couldn't check: a candidate who clears every round on video, passes every check, and still isn't the person who shows up on day one. Since then, the replies have told us we weren't exaggerating. Recruiters messaging us about the interview that "felt off." A SOC lead who'd lived through a KnowBe4-style scare. HR leaders quietly admitting they have no idea whether the face in round one matched the face in round four — because nothing in their stack checks. That's the gap. And it's the whole reason ContentLens exists. So if you've followed this series, here's the one thing worth doing now: → Ask your team a single question — how would we know if a candidate was deepfaked or proxied through our process today? If the honest answer is "we wouldn't," you're not behind. You're where most teams are. The difference is whether you close the gap before it costs you a quarter, a role, or a breach. ContentLens sits inside your existing interview stack — a bot joins the call, flags face swaps and voice clones in real time, catches whispered coaching through an earpiece, and spots candidates reading AI-generated answers off a second screen. It also verifies the same person shows up across every round and on joining day. No new workflow. No forensic training. Just the check that was always missing. If this series put the problem on your radar, let's talk. A 20-minute walkthrough is usually all it takes to see where your process is exposed. DM open, or comment "verify" and we'll reach out. Rohan Sahu Shivi Mithal #TalentAcquisition #HumanResources #Recruitment #HRTech #DeepfakeDetection #HiringFraud #ContentLens

A candidate can clear four rounds of interviews, pass every background check, and your company can still onboard someone who was never on the call. We wrote about that in our last two posts — proxy interviewing, real-time face swaps, a real identity with a different person behind it. The uncomfortable part wasn't that it's possible. It's that nothing in a normal hiring stack is positioned to notice. So, we built what your hiring stack can't see — and kept it out of the way. Here's how it actually works: → A recruiter invites the ContentLens bot into the interview, like any other participant. → It monitors the candidate's video and audio through the call. → If a face swap or voice clone is present, it's flagged in real time — the recruiter and interviewer are warned during the call, not in a post-mortem. → Every result is viewable and auditable later, per candidate. → And it checks the one thing no one else does: that the same face and voice show up across every round — and on joining day. That last point is the whole thing. Verifying an identity and verifying the person are two different problems. Most hiring processes only solve the first. This solves the second — without asking your team to become forensic analysts. HR and TA leaders: is this something your process accounts for today? Genuinely curious where people are on this. Rohan Sahu | Shivi Mithal | Pratyush Karna #TalentAcquisition #HumanResources #Recruitment #HRTech #DeepfakeDetection #HiringFraud #ContentLens

In our previous post, we brought out how a candidate clears multiple rounds of interview using deepfakes; manager sees candidate struggle with the simplest tasks post joining. In this post, we want to bring out how deep the scam really goes. "IT Interview Proxy Service" is available in many geographies — listed on mainstream B2B marketplaces, priced per interview, specialized by tech stack. Some services even bundle "proxy job support" so the same expert does the work remotely while the candidate sits at the desk. A more qualified person clears the rounds for candidate on video. Candidate joins on day one. The company never knows. Same identity. Same resume. Same name. Just not the same person who was on camera. The deception is scoped to the interview — the one stage of hiring nobody verifies. That's the everyday version. The extreme version looks different. In July 2024, KnowBe4 — a US cybersecurity company — discovered that a newly hired Principal Software Engineer wasn't who they'd interviewed. The identity was real, stolen from someone else. The face on the calls was AI-enhanced. Detection didn't come from HR. It came from their SOC, twenty-five minutes after the laptop arrived — when malware started loading onto it. Different motives. Same technique. Same blind spot. → The candidate passes live video rounds → The background check is clean because the identity is clean → The interview stack — webcam, Zoom, human judgment — wasn't built to detect a real-time face swap The lesson isn't "do better background checks." It's that verifying an identity and verifying the person are two different things — and most hiring processes only do the first. That's the gap ContentLens closes. Had a "something felt off" moment in an interview recently? DM open. Rohan Sahu | Shivi Mithal #TalentAcquisition #HumanResources #Recruitment #HRTech #DeepfakeDetection #HiringFraud #ContentLens

A senior engineer aces three rounds of video interviews. Strong resume, confident answers, glowing references. They get the offer. On day one, the person who logs in behaves differently from the one you interviewed — and within days, hiring manager is wondering how this person cleared interview rounds or malware is on your network. This isn't hypothetical. It's already happened to companies that ran reference calls, background checks, and multiple video rounds. Deepfake candidates are quietly becoming one of the biggest blind spots in hiring: → Gartner predicts 1 in 4 candidate profiles will be fake by 2028 → 6% of job candidates admit to participating in interview fraud — posing as someone else or having someone else pose for them (Gartner survey of 3,000 candidates, 2025) → 62% of HR pros believe candidates are now better at faking identity with AI than teams are at spotting it (Checkr Hiring Hoax Survey, 2025) → 31% of companies have already deployed AI or deepfake detection software (StudyFinds survey of 874 hiring professionals, 2025) The tools are cheap, the methods are improving fast, and remote-first hiring gives fraudsters a controlled environment to operate in. ContentLens was built for exactly this moment. We help HR and TA teams detect: → Face-swap and real-time video manipulation in live interviews → Voice cloning in phone and video screens → AI-generated and altered media in candidate submissions Think of it as the verification layer your interview stack is missing — quietly working in the background so your team can focus on hiring great people, not unmasking fake ones. HR and TA leaders: if this is on your radar (or should be), I'd love to compare notes. Rohan Sahu | Shivi Mithal #TalentAcquisition #HumanResources #Recruitment #HRTech #DeepfakeDetection #HiringFraud #ContentLens

We are excited to welcome Shivi Mithal as Strategic Advisor for Cybersecurity to ContentLens. Shivi brings nearly three decades of experience across software development, consulting, and delivery leadership — with deep roots in cybersecurity. His hands-on expertise spans email security, deepfake protection, insider risk management, and security awareness training, with a strong background in product management within the cybersecurity space. He has led and scaled large technology organisations and has been a recognised voice in the India technology leadership community, including being honoured at the GBS Leadership Awards 2025 and GCC Leadership Conclave 2026. At ContentLens, we're building tools that power authenticity in the digital world — and cybersecurity is a domain where that mission matters most. Deepfakes, AI-generated content, and digital impersonation are no longer edge cases; they're frontline threats. Shivi will work closely with us to fine-tune our products for the specific needs of cybersecurity use cases, helping us sharpen our offering for enterprise and security-first buyers and grow the cybersecurity business. We believe the best advisors are those who've operated in the same trenches as our clients. Shivi brings exactly that — and we're excited to grow together. Welcome aboard, Shivi! 🙌 #Cybersecurity, #DeepfakeDetection, #ContentAuthenticity, #ResponsibleAI, #ContentLensAI

[𝐈𝐧𝐝𝐢𝐚 𝐝𝐨𝐮𝐛𝐥𝐞𝐬 𝐝𝐨𝐰𝐧 𝐨𝐧 𝐜𝐨𝐧𝐭𝐞𝐧𝐭 𝐚𝐮𝐭𝐡𝐞𝐧𝐭𝐢𝐜𝐢𝐭𝐲] PM Modi closes his address at the India AI Impact Summit with a call for global standards to address the menace of deepfakes, misinformation by implementing global standards for content authenticity with specific mention of 𝒘𝒂𝒕𝒆𝒓𝒎𝒂𝒓𝒌𝒊𝒏𝒈 & 𝒄𝒐𝒏𝒕𝒆𝒏𝒕 𝒍𝒂𝒃𝒆𝒍𝒍𝒊𝒏𝒈 (similar to "nutritional labels"). The leading global interoperable standard for this today is the Coalition for Content Provenance and Authenticity (C2PA) and ContentLens is proud to be a contributing member of C2PA, the first company in India with indigenous technology implemented globally, alongside industry giants such as Google DeepMind Adobe Samsung Research OpenAI Microsoft etc. Credit to C2PA steering members & leaders (Andy Parsons, Leonard Rosenthol, Sherif Hanna, Yogesh Deshpande Scott Perry etc.) for driving alignment amongst companies across countries & industries. To the Content Authenticity Initiative for driving awareness and engaging with stakeholders interested in content authenticity. DM or email us (team@contentlens.ai ) to know more about our products and ongoing discussions with Ministry of Electronics and Information Technology (MEITY) towards content authenticity, cybersecurity, identity verification etc. cc: Rohan Sahu Dhruv Suri Daksh Ramesh Chawla Pratyush Karna Special mention to our advisors 1. David Owczarek (from Catamount Music) for also being an early adopter & advocate of our products in the music industry, leading the way in protecting original music through imperceptible watermarking before it is shared publicly. 2. Jagabandhu Mishra for voice authentication and speaker verification related R&D PM's speech https://lnkd.in/gEaYWj_Y (watch from 1:20 onwards)