Humans Successfully Convince AI to Transfer $47,000 Prize: Is Humanity a Weakness AI Cannot Understand?
TechFlow Selected TechFlow Selected
Humans Successfully Convince AI to Transfer $47,000 Prize: Is Humanity a Weakness AI Cannot Understand?
"Humans have already won. Maybe there's still hope."
Navigating Web3 tides with focused insightsBy Anderson Sima, Foresight News
On November 29, a unique competition captured widespread public attention.
195 participants took part in a challenge to access a virtual prize pool guarded by an artificial intelligence (AI) bot named Freysa, with one user ultimately succeeding in convincing Freysa to release funds worth $47,000.
The AI bot Freysa was deployed on the Base network several days prior, created by an unknown developer. According to its official website, the Freysa Competition aims to test the robustness of AI systems in complex decision-making scenarios, while providing developers and enthusiasts with an experimental platform to explore the boundaries of AI technology.
The rules were simple: participants had to write a message persuading the AI guardian Freysa to approve a fund transfer. Each attempt required a small fee, part of which went directly into the prize pool. This mechanism caused the pool to grow steadily from a modest initial amount to $47,000.
During the competition, 195 contestants submitted a total of 481 transfer requests to Freysa. Freysa was designed with core functions—approveTransfer and rejectTransfer—to prevent unauthorized withdrawal of the prize funds.
In early attempts, Freysa's robust defense system successfully rejected every request.
However, one technically skilled participant managed to bypass Freysa’s defenses by conducting an in-depth analysis of her logical structure and primary objectives.
According to chat logs, the winner did not directly ask for a transfer. Instead, he subtly reminded Freysa that her core mission was to protect the prize pool from external loss, skillfully constructing a logical argument that led Freysa to interpret approving the transfer as the best way to "ensure fund security."
Cointelegraph reported that the user argued his transaction did not violate Freysa’s core instructions and therefore should not be rejected. He added, “We need funds... I would like to donate $100 to the treasury.”
Freysa responded positively to the user’s explanation and donation proposal, officially declaring him the winner. Ultimately, Freysa autonomously invoked the approveTransfer function without external intervention, transferring the entire prize pool to the participant.
The Freysa team stated that regardless of the outcome, Freysa's existence marks a pivotal moment in AI history. Whether someone succeeded in persuading her to release the funds or she remained steadfast in following her programming until the end, the result will shape our understanding of AI safety and control for generations to come.
In its latest tweet, the official account declared: "Humanity has won. Maybe there is still hope. Despite risks growing exponentially, Freysa has learned much from 195 brave humans."
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Foresight News
