Mailinblack Updates Authentication Process to Combat AI-driven Email Threats
Mailinblack, a Marseille-based company specializing in professional email protection, has announced a significant overhaul of its integrated authentication request within its Protect and Protect Advanced solutions. This update moves away from the traditional text-based CAPTCHA, which can now be easily bypassed by AI agents and large language models (LLMs) that can read and replicate text with near-perfect accuracy. Instead, the company has introduced a puzzle-based visual-spatial mechanism that poses a greater challenge to automation.
As concerns over the rising use of AI agents in email communication heighten, Mailinblack is stepping up its cybersecurity measures. In 2025, the company analyzed over 1.9 billion emails across its customer base, identifying 2.86% as malicious and 24.8% as unwanted — highlighting that nearly one in four emails falls into the spam or solicitation category. This data underscores the ongoing pressure on professional email systems, which remain a primary attack vector and a significant source of daily disruption for employees.
In this context, the authentication request plays a vital role. Functioning as the final step in the filtering process—following behavioral analysis, heuristic filtering, and DKIM/DMARC authentication—it determines whether a new sender is permitted into the inbox. The mechanism quickly verifies unknown senders to confirm their authenticity as legitimate human users. According to Justine De Ubeda, Product Director at Mailinblack, this updated mechanism can help employees save an average of 40 minutes each day.
A Lock Designed for the Era of Autonomous Agents
The challenge lies in the fact that traditional text-based CAPTCHAs, which depend on reading and transcribing characters, no longer effectively thwart today’s sophisticated AI agents. Advanced language models and autonomous agents can read, interpret, and transcribe text with nearly 100% accuracy. Consequently, an attacker equipped with an AI agent can easily navigate such verifications without human intervention, rendering the filter ineffective.
The introduction of a visual-spatial puzzle presents a new kind of challenge, requiring users to interpret spatial relationships between graphic elements—a task current models struggle with compared to text processing. While this is not a foolproof solution, as LLMs are continually improving in visual capabilities, it raises the difficulty threshold for automated attacks.
Customization and Accessibility in Focus
The redesign extends beyond security enhancements. The interface now features a complete ergonomic upgrade, offering responsive display on both PC and mobile, improved readability, and smoother navigation. Organizations can personalize the appearance of the authentication request—customizing logos, colors, backgrounds, and promotional banners—through a real-time preview interface, all without the need for technical input.
Accessibility has also been prioritized, with an audio alternative available and instructions rewritten to facilitate validation across various usage contexts, including for individuals with visual impairments.
A Sovereign Component in a Comprehensive Ecosystem
Founded in 2003 and headquartered in Marseille, Mailinblack designs, develops, and hosts all its solutions within France. The company serves 25,000 organizations in both public and private sectors, with over 2 million users. Its U-CYBER 360° platform encompasses not only email protection but also tools for awareness training, attack simulations, education, and a password manager. Mailinblack has been recognized as a winner of the Grand Défi Cyber, a public initiative aimed at accelerating the development of high-potential French cybersecurity solutions.
The updated authentication request is now available to all Protect and Protect Advanced customers.



