Google has expanded its bug bounty programme for vulnerabilities in generative AI systems

Jakub Porzycki/Getty Images

Google has supplemented its rewards programme for finding vulnerabilities with attack scenarios specific to generative artificial intelligence.

Here's What We Know

In the bug bounty programme, Google pays ethical hackers cash rewards for finding and responsibly disclosing vulnerabilities. According to the company, this incentivises research into AI security and will help identify potential problems to make the technology more secure.

Given the new security concerns due to AI, Google has decided to rethink how it categorises bugs in its systems. To do this, the company is using data from its AI Red Team modelling various cyberattacks.

In particular, the team discovered the vulnerabilities of language models to the introduction of malicious hints and the extraction of training data by attackers. Such scenarios are now covered by the bug bounty programme.

However, bugs discovered that involve copyright infringement or the extraction of insensitive data will not be rewarded.

The amount of the reward will depend on the severity of the vulnerability found. Google paid researchers more than $12 million in 2022.

Source: TechCrunch