As time passes, digital scams become as formidable as the traditional. This is why on Thursday, May 8, Google lifted the veil on a set of new countermeasures based on AI.
These are intended to strengthen the fight against online scams on Chrome, Search (the search engine) and Android. The company provides in particular to integrate Gemini Nano into the Chrome version 137 For office computers.
In case you do not know, Gemini Nano is the house model (LLM) of the house designed to operate directly on the devices. The purpose of integration is to strengthen the capacity of Safe Browsing.
But why Google boosts Safe Browsing at AI?
According to Google, this approach instantly analyzes the websites visited. She will even go so far as to anticipate unprecedented scam attempts.
Thanks to the ability of Gemini Nano To analyze the varied and often misleading structure of online content, security systems can adapt more quickly to the new techniques used by cybercriminals.
“” The LLM is particularly suitable for this type of task, because it is able to interpret subtle and changing signals on web pages, which allows us to quickly react to threats in constant evolution “Said the company.
This technology is already used to counter fraud linked to Distance technical support. The type of scam that pushes users to believe that a dysfunction affects their device in order to encourage them to share sensitive information.
How will it work? Basically, Gemini Nano analyzes web pages looking for suspect elements. He then transmits the signals to Safe Browsing, which assesses if the page really represents a risk.
And of course, precautions have been set up to guarantee effective and secure execution. Engineers Jasika Bawa,, Andy Lim And Xinghui Lu From the Google Chrome security team detailed them:
“” The model is only activated in rare occasions and works completely locally, without transferring data. We also optimize the consumption of resources by limiting the number of tokens treated, by performing the process asynchronously to avoid disturbing navigation, and by applying quotas to Supervise the use of the GPU. »»
Google does not intend to stop there
This is only the beginning indeed. Google plans to expand the use of this technology to other very widespread scams. Including those that usurp Delivery services or unpaid tolls.
This reinforced feature will also be deployed on Chrome for Android by the end of the year. And in parallel, the detection systems of scams integrated into the Google search engine have also been largely improved.
Thanks to AI, these tools are now able to identify twenty times more fraudulent pages than before, excluding this content results.
According to the firm, this advance has already reduced Over 80 % Scams pretending to be customer airline services. And Over 70 % Those imitating official sites related to visas or government services – just in the year 2024.
Obviously, these updates are not only made to protect us on the internet. They are also part of Google’s efforts to integrate AI in all its ecosystem.
So what do you say?
.