How to Make AI Pictures of Yourself?

Why Everyone's Making AI Pictures of Themselves (and Why Experts Are Warning About It)

In late 2022, Lensa's "magic avatars" swept social media, hitting No. 1 in the iOS App Store's Photo & Video category and racking up more than five million downloads in a single month. The trend hasn't slowed: roughly 15.5 billion AI-generated images have been created to date, with about 34 million new ones appearing every day, according to industry analysis. The market itself is projected to grow from USD 8.7 billion in 2024 to USD 60.8 billion by 2030, driven largely by demand for personalized visuals.

The appeal is obvious. Upload a handful of photos, pay a few dollars — Lensa's Magic Avatars cost $7.99 and require at least 10 images — and receive polished, stylized portraits of yourself in minutes. Digital twin tools promise the same speed, cloning your likeness into a personalized avatar almost instantly.

But cybersecurity experts say most users have no idea what they're actually handing over. "We don't know what they're going to do with that data," AI and cybersecurity specialist Mari Galloway told TODAY.com. "How are they deleting it? How is the data encrypted? We don't really know those details because they don't really share that information with us."

The core problem is that deletion claims are unverifiable. Lensa's privacy policy states Face Data is automatically deleted within 24 hours, and Prisma Labs says photos are permanently erased once avatars are generated — but the company did not respond to TODAY's request for comment, and computer scientist Juergen Schmidhuber openly questions how such deletion works when removing data from established platforms is notoriously difficult.

Experts also flag risks most users never consider:

• Deleting the app alone doesn't delete your data — Galloway warns the information is only removed when you delete your account.
• Personal details collected at signup, like your name and email, could be used for tracking or, if exposed, potentially to open fake bank accounts or credit cards.
• If the app changes ownership, any remaining data transfers to the new owner.
• Facial-photo apps could make it easier for bad actors to find people, according to NBC News cybersecurity experts.

This is why a safety-first mindset matters before you upload your face anywhere. Read the privacy policy, limit the personal data you share, and verify what the tool actually does with your images. And as with any AI tool, the same scrutiny applies to the software ecosystem around it — which is exactly why SkillzBench is building security scanning, version pinning, and sandbox benchmarking for agent skills, with a free skills directory and MCP connector on the way.

The popularity of AI self-portraits is real and growing. So are the risks — and understanding both is the first step to participating in the trend safely.

How AI Self-Portrait Tools Actually Work (and What You're Really Uploading)

You upload ten selfies, wait a few minutes, and a polished digital you stares back from the screen. Behind that magic sits a surprisingly simple pipeline — and a data trail most users never think about.

The technology itself comes down to three model families. Research on generative image tools identifies GANs, VAEs, and diffusion models as the engines that "generate realistic images based on written descriptions." GANs in particular use two competing networks to produce results that are, per industry analysis, "nearly indistinguishable from real photos."

The typical workflow is consistent across tools. Digital twin platforms have users upload a photo or video and return a personalized AI avatar in minutes. Lensa AI, which hit No. 1 in the iOS App Store's Photo & Video category, requires at least 10 photos, assigns numerical values to facial points, and combines them into new portraits, according to expert reporting.

But here's what you're really handing over:

• Your face data — biometric identifiers that tools like Lensa convert into numerical facial measurements
• Your email address and account details, required at signup
• Payment information — Lensa's Magic Avatars cost $7.99, and headshot tools run $29–$69 per person
• Whatever remains on servers after generation, since deletion claims can't be independently verified

That last point matters more than most users realize. Cybersecurity specialist Mari Galloway told TODAY that users don't know how data is encrypted or deleted: "We don't really know those details because they don't really share that information with us." She also warned that if a company changes ownership, your remaining data transfers to the new owner — and could theoretically be used to open fake accounts.

Computer scientist Juergen Schmidhuber raised a sharper question: how does Lensa delete data so easily when deleting data from platforms like Facebook is "very, very difficult"? Lensa's privacy policy claims face data is auto-deleted within 24 hours, but the company did not respond to requests for comment when pressed.

The practical takeaway: deleting the app is not deleting your data — you must delete the account itself. And if you want to automate the surrounding workflow safely, agent skills can help with batch image handling and publishing; SkillzBench is building a free directory with security scanning and benchmarking so you can verify which skills are safe before they touch your files.

The Safety Playbook: Protecting Your Face and Data Before You Upload

Your face is the one password you can never reset, so think carefully before handing it to an app. The AI self-portrait trend is enormous — Lensa alone hit No. 1 in the iOS App Store's Photo & Video category — but cybersecurity experts say the risks deserve as much attention as the results.

Start by reading the privacy policy, and treat deletion promises with healthy skepticism. Lensa claims your Face Data is deleted within 24 hours, yet AI and cybersecurity specialist Mari Galloway points out: "We don't know what they're going to do with that data... How are they deleting it? How is the data encrypted? We don't really know those details because they don't really share that information with us." Computer scientist Juergen Schmidhuber raises the same question — if deleting data from platforms like Facebook is "very, very difficult," why is it so easy here?

Second, understand that deleting the app is not deleting your data. Galloway is blunt: "You only get rid of that information when you delete your account... just deleting the app doesn't mean that they're not still storing that information on you." She also warns that if the company changes ownership, whatever data remains transfers to the new owner.

Before you upload anything, run through this checklist:

• Read the privacy policy and look specifically for how long photos are retained, whether they train models on your face, and who else they share data with.
• Delete your account — not just the app — when you're finished with the tool.
• Limit what you share at signup: use an email address you don't mind exposing, and skip tools that demand more personal data than the job requires.
• Prefer tools that state SOC 2 and GDPR compliance — some enterprise avatar platforms, like those surveyed in this comparison of AI avatar tools, advertise these certifications for a reason.

The stakes are real. Galloway notes that exposed personal data — name, email, address — could be used for tracking or, in a worst case, to open fake bank accounts or credit cards. And NBC News cybersecurity experts warned that facial-photo apps "could make it easier to find people."

The same verify-before-you-trust habit applies to the AI tooling you use at work. SkillzBench is building a free directory that security-scans, version-pins, and sandbox-benchmarks agent skills before you run them — the same due diligence you should apply to any tool that touches your data. Your likeness is worth protecting, and a few minutes of caution before upload saves far more trouble later.

Getting Results That Look Like You: Quality, Bias, and Disclosure

You've generated your AI portraits, paid your fee, and scrolled through the results — and now comes the moment of truth. Not every image that comes back is one you should actually use, and some may not even look like you. Before any AI headshot goes on a LinkedIn profile or company bio, it needs a careful quality check.

The most common failures are easy to spot once you know what to look for. Practitioners reviewing AI headshots report recurring problems including eye and lighting artifacts, fake teeth, waxy skin, hair inconsistencies, and the "doesn't look like me" problem, according to industry analysis of AI headshot tools. A survey of 1,600 participants found that 38% described AI headshots as "soulless" — a reminder that technical polish doesn't guarantee a convincing likeness.

Bias is a second, more serious concern. When MIT graduate Rona Wang ran her photo through an AI tool, it lightened her skin and added blue eyes, prompting her to ask the Boston Globe: "does this thing think I should become white to become more professional?" A Bloomberg test of Stable Diffusion across 17 occupations found that more than 80% of "inmate" images had the darkest skin tones, and women appeared in only 3% of "judge" images despite making up 34% of real judges. Earlier MIT Gender Shades research measured 99%+ accuracy for white men versus 65.3% for darker-skinned women. The lesson: check that the output actually represents you — your skin tone, your features, your age — not a stereotype.

Disclosure is the third gate, and it matters more than many users assume. In a Ringover survey of 1,087 recruiters, 76.5% actually preferred AI headshots in blind comparisons — but 88% say AI-generated images should be disclosed, and 66% would be put off when they discover one was hidden. A separate Getty Images survey found 90% of adults want to know whether an image is AI-created. In other words, the image itself can pass inspection; the concealment is what damages trust. Emerging rules like California's AB-723 disclosure requirements are also turning this from an etiquette question into a compliance one.

A practical pre-use checklist:

• Zoom in on teeth, eyes, hair, and skin texture — regenerate anything with obvious artifacts.
• Compare the result side-by-side with a real photo and ask whether a colleague would recognize you.
• Watch for bias: skin lightening, age shifts, or styling that doesn't match how you actually present.
• Label AI-generated portraits clearly in professional contexts, rather than letting viewers guess.

If you're producing portraits at scale — for a team page or a batch of profile updates — agent skills can automate the rest of the workflow, from batch image handling to publishing. SkillzBench is building a free directory where you can browse, security-scan, and benchmark those skills, with its waitlist currently live.

Your Step-by-Step Workflow: From Photo Upload to Polished, Safe AI Portrait

Creating an AI picture of yourself takes minutes — but doing it safely takes a little more care. The workflow is simple: pick a tool, upload photos, generate, then vet and share responsibly. Here's how to run each step without giving away more of yourself than you intend.

Step 1: Choose your tool. Headshot apps like Lensa require at least 10 photos and convert facial features into numerical data points to generate new portraits, while digital-twin tools like D-ID can animate a single uploaded photo into a talking avatar "in minutes." Before committing, check for stated compliance credentials — some avatar platforms, such as Synthesia and Colossyan, advertise SOC 2 and GDPR compliance.

Step 2: Read the privacy policy first. Cybersecurity specialist Mari Galloway warns that users can't verify how these apps delete, encrypt, or use uploaded data: "We don't know what they're going to do with that data." Lensa's policy claims Face Data is deleted within 24 hours, but the company declined to comment when asked, so treat deletion claims as unverified. Limit the personal data you share at signup — an exposed email can be used for tracking.

Step 3: Upload and generate. Provide the required number of clear, well-lit photos, pay any fee (Lensa's Magic Avatars run $7.99), and wait for results.

Step 4: Vet every output. Common AI failures include waxy skin, fake teeth, eye artifacts, and the "doesn't look like me" problem. Bias is also documented: an MIT study found facial analysis was 99%+ accurate for white men but only 65.3% accurate for darker-skinned women, and one MIT graduate found an AI tool lightened her skin and added blue eyes. Reject anything that misrepresents you.

Step 5: Post responsibly. Disclosure matters — 88% of recruiters say AI headshot use should be disclosed, and 66% are put off when they discover it after the fact. And when you're done, delete your account, not just the app — as Galloway puts it, removing the app alone doesn't stop the company from storing your information.

Once your portraits are ready, agent skills can automate the downstream busywork — batch-resizing images, publishing to profiles, scheduling posts. Because roughly 26% of published agent skills carry at least one vulnerability, it pays to vet them first: SkillzBench is building a free directory with security scanning, version pinning, and sandbox benchmarking, plus a free MCP connector to load only the skills you choose. The waitlist is live now.

• Prefer tools with documented compliance certifications (SOC 2, GDPR)
• Share the minimum personal data required at signup
• Inspect outputs for artifacts and bias before using them
• Disclose AI-generated images in professional contexts
• Delete your account — not just the app — when finished

Done well, an AI portrait is a fast, affordable upgrade — professional headshots boost perceived competence by nearly 76% — as long as you stay in control of your data at every step.