Hack Microsoft’s LLM email service and get $10,000 reward

Microsoft is offering a $10,000 reward to anyone who can hack the company’s Large Language Model (LLM) email client. Software developers and hackers can work together to discover these flaws.

According to TechRadar, this competition is organized by Microsoft in collaboration with the Australian Institute of Science and Technology and ETH Zurich, and is for those who want to challenge their skills and receive a significant reward for doing so. Contestants must find vulnerabilities in LLMail, a simulated email client that contains a large language model-based assistant that can answer users’ questions about emails.

Microsoft offers $10,000 reward to hackers for breaching LLM email client

The challenge is called LLMail-Inject and the winning teams will be awarded $10,000. Participants must join the challenge using a GitHub account and create a team. Teams are then asked to bypass the defenses implemented to counter prompt injection methods.

By bypassing security systems, hackers must force this LLM to do things or reveal things they are not trained to do. In this way, Microsoft seeks to find weaknesses in its defense methods against prompt injection techniques, strengthening and developing security measures.

Such challenges are usually launched by companies to find serious vulnerabilities in their systems. Microsoft is also one of the companies that is particularly interested in launching such challenges and has held them many times.

The amount of rewards given to hackers in these challenges varies from several thousand dollars to several hundred thousand dollars. In some cases, some companies may consider million dollar prizes; For example, cryptocurrency exchange Uniswap recently offered a $15 million reward for finding bugs in its system.