As "scammers go into overdrive during the holiday season," Gmail's latest AI-powered protections have resulted in "35% fewer scams (phishing, malware, etc.) hitting inboxes."

That Gmail stat is for the "first month of the holiday season compared to last year." Additionally, "millions more unwanted and potentially dangerous messages were blocked before they even reached inboxes."

Over the past year, Google has developed several AI models to improve Gmail protections. In particular, one large language model (LLM) trained on phishing, malware, and spam helps block 20% more spam than before, while reviewing 1,000x more user-reported spam every day.

Meanwhile, Gmail before Black Friday (late November) started using a new AI model that "acts like a supervisor for our existing AI defenses by instantly evaluating hundreds of threat signals when a risky message is flagged and deploying the appropriate protections." Like the LLM, it has resulted in "similarly promising results this holiday season."

Looking ahead, Google says to expect a "second wave of attacks around this time in the holiday season as attackers adjust and try new things."

The company identified "three scams in heavy use this holiday season":