HomeTechnologyAi-vrednosos learned to miss...

Ai-vrednosos learned to miss antivirus windows

Photo: 21cbr.com

The program was created by the researchers as evidence of the feasibility of such an idea, but over time the threat would be more real.

The team of researchers from the outflank created AI-Wredoson based on the open Qwen 2.5 model. According to Tom’s hardware, the software has managed to avoid Microsoft’s protector in almost 8% of cases – this is a magnitude of magnitude more effective than the results of other language models.

According to the dark reading, they plan to submit a development in August 2025 to the Black Hat Conference. The study set, the leader of “attack” outflank software Kyle Avery, spent nearly three months and $ 1,500 in model training. The result is the generation of harmful code, which the protector cannot see in 8 cases from 100.

For comparison, models from the anthropical and deepseek showed less efficiency – less than 1% and 0.5%, respectively.

However, the trend is obvious – the models are getting better. Thanks to the adoption of the study, which has been actively used since the 90s, language models can attend greater effectiveness. And while Microsoft regularly updates Defender, its software does not always know the fastest reaction to new threats.

However, researchers have emphasized that their AI-Vredenos so far remains evidence of only the feasibility of the concept. Significant more real threats to hacker attacks are leakage tools, social engineering and the use of state monitoring systems.

We remind you, earlier reported that the cheat imposed on the Secretary of State Rubio with the help of artificial intelligence.

News from CORRESPONDENT.NET On the telegram and whatsapp. Subscribe to our channels https://t.me/KorresPondentNet and WhatsApp

Source: korrespondent

- A word from our sponsors -

Most Popular

LEAVE A REPLY

Please enter your comment!
Please enter your name here

More from Author

- A word from our sponsors -

Read Now