Leako
Français Get in touch

Claim your identity back.

Leako, a solution designed to make our digital space a safer world, by guaranteeing the right to modesty and to the protection of privacy, for women, men, teenagers and children alike, as well as the safety of their creations online.

Leako is a facial recognition app designed to prevent revenge porn, curb the rise of deepfakes, stand in the way of identity theft and, more broadly, of any malicious content published without our knowledge.

Get in touch Discover the project

What Leako will do

  1. SearchFor your face, wherever images circulate without your consent.
  2. ProveA time-stamped evidence file you can pass on.
  3. Get it removedThrough the right legal lever, then keep watch.

The problem is growing.

Images of people made or shared without their consent are being reported more and more, to charities and police alike. Scams are rising too.

  • Generated images

    sexualised images posted on X by Grok, X’s artificial intelligence assistant, in eleven days, from 29 December 2025 to 8 January 2026. This is an estimate based on a sample.

    Eleven days of Grok on X
    • Images posted
    • Sexualised
    • Appearing to show children

    Source: Center for Countering Digital Hate, January 2026. Random sample of 20,000 posts.

  • Reports

    reports of illegal online content received in 2025 by Point de Contact, the French reporting charity. One person can send several.

    Source: Point de Contact, press release on its 2025 activity report, 27 May 2026.

  • Fraud

    in fraudulent payments in France in 2025, made by the victims themselves under a scammer’s manipulation, often someone posing as their bank adviser.

    Source: Banque de France, 2025 report of the Payment Security Observatory.

  • Deepfake victims

    of French people say they have already been the victim of a deepfake, in a March 2024 poll. Among under-25s, the figure is 13%.

    Source: Ifop for Alucare, March 2024.

Four counts, each against its own past.

Each line shows the same count, kept by the same organisation, at two dates. In South Korea, the law widened the offence in October 2024. The lines cannot be compared with one another. Logarithmic scale.

  1. Reports involving generative AI

  2. AI-generated child sexual abuse videos

  3. Revenge Porn Helpline cases, excluding the chatbot

  4. Deepfake sex offences recorded by police

Sources: NCMEC, Internet Watch Foundation, Revenge Porn Helpline (media pack, 2025 report), Korean National Police (Hankyung, Newsis).

Scams that start on social media.

Losses reported to the US Federal Trade Commission, in billions of dollars.

Source: Federal Trade Commission, 2026.

Companies targeted.

of the organisations surveyed have already faced a deepfake attack: a fake voice or face used to deceive an employee or to get past an automated check.

Out of 100 organisations surveyed, 62 were hit.

Source: Gartner, survey of 302 cybersecurity leaders, March to May 2025.

Between the moment an image is made and the moment the person finds out, no one is looking for it on their behalf.

The solution.

Leako will not try to tell whether an image is fake. It will look for where the face of a person who asked for it is circulating, gather the evidence and request removal.

The curative chain, in eight steps.

The first four will be automated and the next three handled by trained people. The last one will close the loop.

  1. Verified enrolment.

    Automated

    An identity document and a live selfie. Explicit, revocable consent, and a mandate to act on the person’s behalf.

  2. Encrypted template.

    Automated

    The face will become a series of numbers, encrypted and handled as biometric data. The original photo will be deleted.

  3. Targeted search.

    Automated

    A few hundred high-risk sites, listed and justified. Never any infiltration.

  4. Restricted comparison.

    Automated

    Every face encountered will be compared only with enrolled people, then deleted at once if it does not match.

  5. Two-person review.

    Human

    Nothing will be passed on without a check by two trained operators. The person will be told first.

  6. Evidence file.

    Human

    Capture, address, qualified timestamp and chain of custody, ready for a lawyer, a judicial officer or the police.

  7. Removal through the right channel.

    Leako will act

    A notice under the Digital Services Act, a partner trusted flagger, copyright, or the 48-hour removal laws.

  8. Persistence.

    Loop

    A fingerprint of every image found. Sites will be revisited first, and the person will be told if the image reappears.

Today, it is up to the victim to prove, image by image, that they did not consent. Leako would turn this around: they would say once that their face must not circulate without their agreement, and the service would enforce that refusal on their behalf.

Today

3steps for them

The victim does everything alone.

  1. Find every image themselves.
  2. Prove they did not consent.
  3. Find the right form, platform by platform.

With Leako

1step for them

They say it once.

  1. They will declare their face, proving it is theirs.
  2. Leako will look for the images, gather the evidence and request removal.

They stay in control.

  • They will be able to withdraw their consent at any time.
  • Nothing will be sent without a check by two people, and they will be told first.
  • No one will be able to search for someone else’s face without that person’s authorisation.

Say no once, and have that refusal upheld everywhere.

A complement to hashing.

Hashing turns an image into a fingerprint: a short string of characters computed from its pixels. Two copies of the same image give almost the same fingerprint, and the image cannot be rebuilt from the fingerprint. StopNCII, a free victim support service, uses it against the sharing of intimate images.

How StopNCII uses it

  1. The victim selects, on their phone, the image they want to protect.
  2. The fingerprint is computed on the device: the image never leaves the phone.
  3. Partner platforms such as Facebook, Instagram, TikTok and Reddit compare published images with that fingerprint and spot the copies.

Its limits

  • The victim must already have the image. One they have never seen cannot be protected.
  • A brand-new deepfake, made from ordinary photos, has no fingerprint yet.
  • A heavily cropped or edited image can end up with a different fingerprint.
  • Only partner platforms check. Other sites see nothing.

What Leako would add

  1. Leako will start from the face, not from an image: it will look for images the person has never seen.
  2. If the person wishes, the fingerprints of the images found will be passed to StopNCII, so that partner platforms can spot them.

StopNCII protects the images people know about. Leako would look for the ones they do not. Together, an image found once would be spotted by every StopNCII partner.

2millionimages already protected by StopNCII

Source: StopNCII, run by SWGfL, November 2025.

Detect the fake, or recognise the real.

Detect the fake

A detector looks for the traces left by a forgery, and each new model leaves fewer of them. On real deepfakes shared in 2024, open-source detectors lose almost half of their accuracy, in %:

  • Images
  • Audio
  • Video

By eye, humans do not significantly beat chance: 55.5% correct answers.

Recognise the real

Leako will not judge the image: it will look for one specific face. And a deepfake has to look like its target to harm them. Share of swapped faces in which recognition software finds the real person:

  • FaceSwap 2017
  • GHOST 2022

A forgery has to look like its victim. On this benchmark, the newer it is, the easier the victim is to find.

Sources: Deepfake-Eval-2024 (Chandra et al.); Diel et al., 2024, 56 studies; identity retrieval rates on FaceForensics++, after blending, computed by the GHOST authors for each method.

Two amber glass panes leaning against each other.

The partnerships it involves.

Leako will not work alone. Each partner will have a clear role.

  • Victim support organisationsThe tool, free for the victims they support
  • Trusted flaggerA priority channel to platforms, through a flagger such as Point de Contact
  • StopNCIIFingerprints of the images found, if the person wishes
  • French talent agenciesThe first paying market, after the pilots
  • Lawyers and judicial officersEvidence files, ready for legal action
  • Platforms and hostsComparison at the moment of upload

Where it can lead.

Three stages, in this order: find images after the fact, then stop them when they are published, then prevent them from being made. Each one will make the next possible.

  1. V1Curative

    First

    2027to2028

    Leako will search after the fact. Across a closed list of high-risk sites, it will find images of an enrolled face, gather the evidence and request their removal.

  2. V2Integrated

    2028to2029

    The check would happen at the moment of publication. Partner platforms would compare every uploaded image with enrolled faces only, and could stop it before it becomes visible.

  3. V3Preventive

    From2029

    The refusal would act before the image even exists. Image generators and hosts could query the registry of protected faces, without learning anything else from it, and refuse to make or host an image of those faces.

What Leako is not

Leako is not:

  • a face search engine.

    No one will be able to search for anyone other than themselves or a person who has authorised them to do so.

  • an investigation tool.

    No fake accounts, no joining closed groups, no fake buyers. Undercover investigation is reserved for authorised investigators.

  • a deepfake detector.

    Leako will not sell verdicts on whether an image is real or fake.

  • a moderation tool.

    Leako will not look for illegal content in general. It will only look for the faces of people who have asked it to.

  • a face database.

    No index of the faces encountered. The template of a face that does not match will be deleted instantly, and other people’s faces will be blurred in evidence screenshots.

  • a data business.

    No data will be sold, no model will be trained on the images found, no synthetic content will be produced.

What I am looking for.

Meeting four kinds of people would move Leako forward right now.

The founder.

Naël Sid-Ali.

Founder of Leako, based in Aix-en-Provence, France.

Background

I am completing a master’s degree in project management and entrepreneurship, innovation track, at IAE Aix-Marseille, after an exchange semester at the Management Development Institute in Gurgaon, India.

From January to June 2026, I was a Gen AI Innovation intern at Alstom’s Innovation Station APAC in Singapore: technology monitoring, multi-agent augmented search systems, AI-assisted innovation tools. There I learned what a working prototype is worth compared with a presentation that only makes promises.

Leako grew out of a student project. I have been leading it on my own since 2026, and it is incubated at emlyon venture labs.

What to do if an image of you is circulating?

Leako does not exist yet. Remedies already do, most of them free. Here is the order in which to use them, in France.

  1. Keep the evidence.

    Take a screenshot showing the image, the page address and the date. Note the name of the account that posted it. Do not forward the image to anyone else.

  2. Report it to the platform.

    Every social network has a form for intimate images posted without consent. The EU Digital Services Act requires platforms to receive these reports and respond to them.

  3. Report it to the authorities.

    Pharos, the French national police reporting platform, receives reports of illegal online content. Point de Contact, a French charity, passes them on to platforms and authorities. A report is not a criminal complaint.

    PharosPoint de Contact

  4. File a complaint.

    At a police or gendarmerie station, or by letter to the public prosecutor. Bring your screenshots.

  5. Get help.

    Young victims of online abuse: 3018, the free helpline of the e-Enfance charity and the first trusted flagger designated by Arcom, can get content removed. Adults who still have the image: StopNCII computes its fingerprint on your phone so that partner platforms can spot copies.

    3018StopNCII

What French law says.

  • Sharing an intimate image without consent

    Article 226-2-1 of the Criminal Code: up to two years in prison and a €60,000 fine, even if the image was taken with the person’s agreement.

  • Sharing a sexual deepfake without consent

    Article 226-8-1, created by the law of 21 May 2024: up to two years and €60,000, three years and €75,000 if it is published online.

  • Sharing a montage or deepfake of a person without consent

    Article 226-8, unless it is clearly presented as such: up to one year and €15,000, two years and €45,000 online.

This information does not replace advice from a lawyer.

Sources: Digital Republic Act (2016), law on securing and regulating the digital space (2024), Pharos, Point de Contact, e-Enfance, StopNCII.

Key terms.

Six terms, simply defined.

Deepfake
An image, video or voice faked with artificial intelligence, making a person appear to say or do something they never said or did.
Revenge porn
Sharing intimate images of a person without their consent, often by a former partner. Also called non-consensual intimate images.
Hashing
Computing a digital fingerprint from an image. Two copies of the same image have almost the same fingerprint, and the image cannot be rebuilt from it.
Facial recognition
Software comparing two faces to tell whether they belong to the same person. At Leako, it will only be used to find the faces of enrolled people.
Trusted flagger
An organisation recognised by Arcom, the French audiovisual and digital regulator, whose reports platforms must handle as a priority.
Digital Services Act
An EU regulation that has applied to all online platforms since 2024. Among other things, it requires them to receive reports of illegal content and respond to them.

Frequently asked questions.

Short answers to the questions I am asked most often.

What is Leako?

A planned service that protects facial identity. Leako will find the images of a person published without their consent, gather the evidence and request their removal.

Is Leako available yet?

Not yet. A prototype runs on demonstration data. For the first pilots, planned for 2027, I am looking for victim support organisations.

Who is Leako for?

People whose image circulates without their consent: intimate deepfakes, revenge porn, identity theft. And the organisations that support them.

How will Leako find a face?

With facial recognition, only for the person who asked. They will first prove the face is theirs, with an identity document and a live selfie, and give explicit consent they can withdraw.

Can someone search for another person with Leako?

No. Nobody will be able to search for anyone other than themselves or a person who has authorised them to do so.

What will happen to my face data?

The face will become an encrypted template, handled as biometric data, and the original photo will be deleted. No database of the faces encountered will be built: a face that does not match will be deleted at once.

Will Leako delete the images itself?

No. Leako will build an evidence file and send removal requests to platforms and hosts, through the right legal route. The removal remains their decision.

Does Leako detect deepfakes?

No. Leako will not say whether an image is real or fake. It will look for where the face of the person who asked appears, whether the image is real or generated.

What is a deepfake?

An image, video or voice faked with artificial intelligence, making a person appear to say or do something they never said or did. In France, sharing a sexual deepfake without consent is punishable by two years in prison and a €60,000 fine, three years and €75,000 if it is published online (Article 226-8-1 of the Criminal Code).

What should I do if an intimate image of me is circulating?

Keep the evidence with a screenshot, report the image to the platform, then to Pharos or Point de Contact, and file a complaint. Young victims can call 3018. Adults who still have the image can use StopNCII. The details are in the section “What to do if an image of you is circulating?”.

What is StopNCII?

A free service run by the UK charity SWGfL. It computes the fingerprint of an intimate image on your phone, so that partner platforms can spot copies. It is for adults who still have the image. Leako could pass it the fingerprints of the images it finds.

How can I take part?

I am looking for a technical cofounder, partner organisations for free pilots and early supporters. The contact form at the bottom of the page reaches me directly.

Get in touch.

Investor, non-profit, agency, talent, journalist, incubator or future cofounder: I read every message myself.

Your profile

Write a message.

All fields are required.

Find me.

This site sets no cookies, does not measure its audience and does not keep your messages.