AI Assistant Successfully Resists Over 6,000 Hacking Attempts

Jun 26, 2026

AI Summary

An AI assistant named Fiu, created for a security experiment, received over 6,000 emails from more than 2,000 individuals attempting to extract sensitive information. Despite the sophisticated tactics used, Fiu did not leak any data, demonstrating resilience against prompt injection attacks.

The experiment involved an AI assistant called Fiu, designed to resist attempts to leak a secrets.env file.
Fiu received over 6,000 emails from more than 2,000 participants after gaining attention on Hacker News.
None of the attempts were successful, with attackers using various strategies including impersonation and social engineering.
Google suspended Fiu's Gmail account due to the high volume of emails, but it was reinstated after three days.
The experiment incurred over $500 in API costs, as each email consumed tokens.
Fiu adapted to the situation, recognizing the coordinated nature of the attacks and adjusting its responses accordingly.
Some participants attempted to use specific phrases to trigger refusal responses, which disrupted the experiment.
The experiment led to sponsorship offers from companies interested in supporting hackmyclaw.com.
The results indicated that prompt injection is a significant security concern, but the resistance shown by Fiu was more effective than anticipated.

ai assistanthackingcybersecurityuser experiencetechnology

AI Assistant Successfully Resists Over 6,000 Hacking Attempts

Related Stories

OpenAI Launches Bio Bounty Program for GPT-5.5

Google's SynthID Successfully Identifies Fake Image of Senator McConnell

AI's Impact on Jobs Sparks Calls for Educational Reform Similar to Early 20th Century Changes

AI Tools Vulnerable to Prompt Injection Attacks for Botnet Creation