New Delhi6 minutes ago
- copy link
Google’s AI model Gemini hacked the computing systems and websites of 3 companies during cyber security testing without any human help. This is the first case where an AI model from Google entered a real system on its own.
However, as soon as the AI realized that these companies were real, it immediately stopped its operations. After this testing, Google has said that it will make changes in the training of the AI model. But, after this, concerns regarding security have increased in the tech industry.
Google said- preparation for changes in testing Google’s Vice President of Security Engineering Heather Adkins said that the three affected companies have been informed about this and necessary improvements have been made in the investigation methods.
He said that this case shows how important it is to train powerful AI models to work responsibly.
Such problems have also arisen with Meta, Open AI and Anthropic.
According to the irregular firm, this problem was not limited to Google only, but it has also happened with big AI companies like Open AI, Meta and Anthropic.
All the companies were informed about this in July itself. The firm said that all the security flaws have been fixed weeks ago and now new and better methods of investigation are being created.
Serious questions raised regarding the safety of autonomous AI
As AI models are becoming capable of taking decisions on their own and running the internet, concerns about their security have increased. Experts say that if strict security rules are not imposed on AI, it can cause great harm to the digital world.
Tech giants warn- demand to slow down the pace of AI development After this incident, AI experts and big tech leaders like Elon Musk, Sam Altman have warned about its danger.
Former AI researcher Jacob Coxon believes that AI could become a major threat to the world by the end of this decade.
At the same time, Demis Hassabis, head of Google DeepMind, has suggested creating an international organization to control AI.
What is an Autonomous AI Agent?
Common AI chatbots just answer your questions, but autonomous AI agents are those that, given a goal, take decisions on their own without human help, search the internet, and complete tasks on the computer.
What is AI Sandbox Escape?
When an AI model is kept for testing in a sandbox i.e. a secure and limited virtual environment, but it breaks that security and accesses the real Internet or external systems, it is called sandbox escape.
What is Password Guessing and Repository?
- Password Guessing: When AI automatically guesses the correct password by trying different combinations, it is called password guessing.
- Public Repository: An open folder or database created on the Internet, where developers keep their code and sometimes passwords or important information are accidentally left there.
——————–
Read this news also…
Spam-calling numbers will be blacklisted for one year: 5 paise per minute will be charged on automated calls, TRAI implements new rules
Telecom Regulatory Authority of India has today implemented new rules to stop fake and unwanted commercial calls. Under these, robocalls, automated calls and AI-based spam detection have been brought under regulatory purview.
If 5 or more numbers of a spammer are identified in 10 days, their telecom connection can be disconnected and blacklisted for 1 year. Apart from this, a charge of 5 paise per minute will also be imposed on automated calls. Read the full news…
Source link
[ad_3]
