AI technology faces new phishing scams and digital watermarking developments

Cybersecurity threats and transparency measures are evolving alongside artificial intelligence. Attackers are utilizing fake websites that impersonate popular AI tools like ChatGPT, Gemini, Claude, and Perplexity to conduct phishing campaigns. These fraudulent sites often pose as AI-driven advertising management tools, prompting users to connect their Google, Meta, TikTok, or Okta accounts. Using a technique known as “browser-in-the-browser” (BitB), scammers display fake login windows that mimic legitimate address bars to steal credentials and bypass multi-factor authentication. These attacks primarily target marketing agency employees and business account administrators.

In response to transparency requirements, OpenAI is planning the gradual introduction of TextGrain, an invisible digital watermark, within the European Union. Unlike traditional visual watermarks, TextGrain alters the statistical patterns of word choices during text generation. This creates a subtle digital trace that specialized software can detect to identify content produced by OpenAI models, even if the text is copied into different applications. While testing has shown results comparable to Google DeepMind’s SynthID, the technology may struggle with short responses or highly constrained content like mathematical explanations.