To combat sophisticated scams driven by artificial intelligence, the Portuguese Association of Direct and Digital Marketing launched Detector de Burla, a free online platform where consumers evaluate suspicious digital communications without needing to create an account.
How the Detector de Burla Platform Operates
Digital fraud continues to evolve as bad actors leverage modern technology to deceive unsuspecting targets. To counter this trend, the Portuguese Association of Direct and Digital Marketing—known as AMD—developed an accessible web tool designed to evaluate suspicious emails, text messages, phone calls, and WhatsApp chats. Users paste or upload questionable content directly into the system, which then examines the text using artificial intelligence models paired with a comprehensive library of known scam patterns.
Rather than relying on isolated keywords, the system looks at holistic context, linguistic patterns, requests for personal or financial data, payment demands, manufactured urgency, and internal inconsistencies. Once the analysis concludes, the platform returns an estimated risk level, breaks down the specific elements that triggered the warning, and provides actionable recommendations. The organization emphasizes that the software acts strictly as an informational aid rather than a definitive legal judgment.
André Novais de Paula, President of AMD, stated via ECO that they do not want to create a false sense of security or replace the authorities, but rather want to help people stop, analyze, and decide with more information.
Data Privacy and Artificial Intelligence Architecture
Because digital scams often involve sensitive personal details, the association built the platform around strict data minimization principles. Users are not required to create an account or provide any personal identification to run an assessment. Before any text reaches the artificial intelligence models via an enterprise API gateway, an automated server-side process masks sensitive items such as IBAN numbers, tax identification numbers (NIF), telephone contacts, credit cards, and email addresses.
While standard identifiers are redacted, URLs remain intact because web addresses are critical for assessing threat levels. Original text submissions are stored temporarily for a maximum period of 30 days to facilitate user re-analyses before being permanently deleted. Furthermore, submitted text is never utilized to train AI models by either the association or its technology providers, and the site remains free of commercial tracking or advertising sponsorships.
Common Digital Scams Targeted by the Tool
The threat library behind the system is continuously updated using emerging patterns derived from user submissions alongside official alerts from cybercrime authorities. Common schemes targeted by the initiative include fraudulent text messages claiming a lost phone, job offers, and impersonation attempts targeting parents under familiar pretexts.
These social engineering tactics typically follow a predictable arc designed to build false trust before convincing victims to surrender funds through direct payments or sharing personal data.
Educational Resources and the Fui Vítima Area
Beyond immediate message scanning, the portal incorporates dedicated educational guides detailing how to spot fraud indicators and recognize recurring scam categories. A specialized section titled “Fui Vítima” provides practical steps for individuals who have already fallen prey to malicious schemes, including direct guidance on contacting relevant financial institutions and regulatory authorities.
The platform also features an aggregate statistical indicators dashboard displaying usage metrics. Prior to its public rollout, organizers previewed the project before a select group of competent consumer protection and cybersecurity entities to establish institutional transparency.
Limitations and Institutional Disclaimers
Creators of the system are transparent about its boundaries. The software cannot guarantee 100% detection accuracy, meaning occasional false positives or false negatives may occur. Organizers reiterate that the tool does not replace judicial authorities, police investigations, or professional legal counsel, nor does it serve as an official criminal reporting channel.
By pairing automated risk scoring with strict privacy safeguards, the initiative aims to foster greater digital literacy across everyday communications without overstating the capabilities of machine learning models in high-stakes fraud detection.
