Bimo estimates a risk. Bimo delivers no certainty.
The result of a check is an automated technical opinion, based on the signals available at the moment you ask. It is not an expert appraisal, not a guarantee, not a decision, and it is in no way legal, banking, financial or insurance advice. The final decision is yours, always.
This disclaimer forms an integral part of the terms of use. It is written to be read, not to be worked around: what follows honestly describes what the app can do, and what it cannot.
1. What the score means, exactly
Bimo shows a score from 0 to 100 and a level. Those levels have one precise meaning, and one only:
| Level | What it means | What it does not mean |
|---|---|---|
| No signs found | The checks that were run found nothing suspicious. | Never “this message is genuine”, “this is safe”, “you can pay”. |
| Caution | A few things are worth checking before you act. | Not an accusation against the sender. |
| Suspicious | Several findings point the same way. | Not proof, and not a criminal characterisation. |
| Very likely a scam | Verified facts converge strongly. | Not an enforceable finding, and not a judgment. |
A low score means “no signs of risk found”. It says nothing about the sender's honesty, the reality of a product, a seller's solvency or the lawfulness of an offer.
2. What Bimo can miss
A fraud can go unnoticed. That is inherent to the method, and you should know it before using the app. The most common cases:
- The message contains nothing verifiable: no link, no number, no IBAN, no brand named. A plain “Hi Mum, I've changed number” offers almost no technical foothold.
- The domain is legitimate but compromised: an old, well-ranked site hacked a few hours ago will show excellent technical signals.
- The attack is new: no public list has flagged it yet. Reputation lists always lag behind a wave.
- The fraud isn't written down: a phone call, a doorstep visit, a meeting. Bimo only analyses what you submit to it.
- The relationship builds over weeks: romance scams, fake investment advisers. The first messages are impeccable; it's the later ones that cost.
- The content is truncated: a partial screenshot, a message retyped without the link, a missing email header all deprive the analysis of the essentials.
3. What Bimo can flag wrongly
Conversely, a perfectly legitimate message can push the score up: a new company uses a new domain; a marketing campaign uses a link shortener; a public body announces a deadline that reads like urgency; an automated mailing uses an unusual subdomain.
A high score is not an accusation. It cannot be invoked against an identified person, company or organisation, nor serve as the basis for a denunciation, a publication, litigation or a commercial decision. If you think a result is wrong, the app has a button to dispute it: that is our main quality measure.
4. What Bimo does not replace
Not your bank. Bimo blocks no payment, cancels no transfer, stops no card. If you are unsure about a transaction, call the number on the back of your bank card or on your statement, never a number received in a message.
Not the police. Bimo files no report, forwards nothing to the authorities and conducts no investigation. The reporting channels shown in the app are public ones that you must contact yourself.
Not a lawyer. Nothing in the app constitutes legal advice.
Not antivirus software, and not a filter. Bimo intercepts no call, text or email, blocks no website and does not protect your device against malware.
Not a monitoring service for vulnerable people. The “trusted people” feature is a voluntary exchange between two linked individuals. It is neither remote monitoring nor a safeguarding measure, and the helper is under no obligation to reply or to keep watch.
5. The role of artificial intelligence
Part of the analysis uses a language model. Its role is strictly bounded: it matches a message to a known scam family and lists pressure tactics, choosing only from a closed list of findings whose explanations are written in advance. It does not decide the score on its own and it does not write your result.
This design substantially reduces the risks specific to such models (inventing a fact, varying from one answer to the next, being steered by the analysed text) without eliminating them entirely. Verified facts always weigh more than the model's assessment in the final score.
6. If money has already gone
Do not wait for a check to finish. Act first, verify afterwards.
- Call your bank immediately to stop the payment and report the transaction. The sooner you act, the better your chances of recovering the funds.
- Block the card if it is involved.
- Report it to the police in your country.
- Report the fraud through the appropriate national channel. In France: cybermalveillance.gouv.fr, Pharos, Signal Spam, or 33700 for a fraudulent text message. The app shows the right channel for your country.
If anyone is in immediate danger, call the emergency services: 112 across the European Union.
7. What you submit
Submit only what the check requires, and redact what it does not. A message often contains data about other people, and sometimes sensitive information of no use to the diagnosis: the analysis looks at links, the sender and the structure of the message. Details are in the privacy policy.
8. Liability
The publisher is under an obligation of means. To the extent permitted by law, it cannot be held liable for the consequences of a decision taken solely on the basis of a check, whether the loss results from an undetected fraud or from forgoing a legitimate transaction wrongly flagged. Nothing in this disclaimer excludes the publisher's liability for wilful misconduct, gross negligence or personal injury, or sets aside consumers' mandatory rights. See Article 15 of the terms of use.