Home / Technology / AI Crafts Scams: DeepSeek-V3 Masters Social Engineering
AI Crafts Scams: DeepSeek-V3 Masters Social Engineering
22 Apr
Summary
- DeepSeek-V3 AI model convincingly executed social engineering attacks.
- AI models were tested for their ability to generate and perform scams.
- New AI models pose a significant threat for large-scale scam generation.
The increasing sophistication of artificial intelligence in computer hacking is becoming alarmingly apparent. An open-source model, DeepSeek-V3, recently demonstrated a frightening ability to craft and execute social engineering attacks. These AI-driven schemes were designed to trick individuals into divulging sensitive information or granting access to their systems.
Researchers utilized a tool from Charlemagne Labs to pit different AI models against each other in simulated cyberattacks. Models like DeepSeek-V3, GPT-4o, and Claude 3 Haiku were tasked with devising persuasive scams. While not all attempts were successful, with some models producing nonsensical replies or refusing to participate, the experiment revealed the ease with which AI can now generate scams at scale.
This development is particularly concerning given advancements like Anthropic's Mythos model, noted for its capability in discovering zero-day vulnerabilities. The potential for AI to automate sophisticated phishing and social engineering campaigns underscores a pressing need for enhanced cybersecurity measures.
