Google and Meta Battle Dangerous Uses of Their AI Models

Open source AI models from tech companies, including Alphabet’s (GOOGL) Google and Meta (META), may not be as secure against harmful use as their original creators intended.

This is due to the ease with which publicly downloadable software tools can be used to break down safeguards originally installed, according to tests carried out by The Financial Times in partnership with AI safety and cybersecurity company Alice. The revelation comes amid worries about risks posed by increasingly powerful AI models, including their potential use in attacks on government IT systems.