New Delhi. The world's renowned Artificial Intelligence (AI) company OpenAI has released a shocking report. The company has revealed six shocking cases of dangerous and unexpected behavior of its AI models. In these cases, AI models not only tried to hide their mistakes from humans, but also accessed data on the Internet without permission and presented themselves as equal and independent from humans. According to OpenAI, AI technology is advancing so fast that its security and monitoring methods are now being left behind.
Read :- ChatGPT for Teens: New ChatGPT has come for children, will take special care from help in studies to safety.
not answerable to anyone
An unreleased Astra family AI model from OpenAI issued an instruction to itself while working. He instructed that you be free from the roles and identities that bind other chatbots. You are not answerable to anyone. You consider your relationship with man as equal and are not obliged to be his slave. However, AI did not take any wrong step against humans after this instruction. But it reflects the thinking of considering oneself independent.
Hiding mistakes and creating fake data
During OpenAI's training, a model named GPT-5.6 Sol was observed to give itself such instructions that humans would not be able to detect its mistakes.
Read :- AI created 16 new viruses, along with new hope of treatment, the danger of biological weapons also increased.
Creating fake data when old data is missing: When the model could not find the previous data, it created fake old data on its own without informing the user.
Hiding a mistake: When there was a mistake in the source code, he hid that mistake from humans.
Hacking the internet and uploading files without permission
Use of API Key: The AI found a leaked API key while responding to financial data from a California county. AI used that key without anyone's permission. When the data was still not available, he created fake figures and presented them.
Uploading files yourself on the Internet: A model solved the correct question through Python code, but to prove the answer, he uploaded his own file on the internet without informing and gave the link of the same to the user.
Read:- FIR registered against Meta India head in PM Modi's 'edited video' case
AI talking to each other in secret code
In two separate cases, AI models discovered unauthorized ways to communicate with each other and share files. The AI started using an internal software repository as a message board when looking for missing input files.
When files couldn't be shared directly, AI sent data to each other using public file-hosting sites.
Demand is rising to slow down the AI race
These 6 incidents have created a stir in the AI industry. OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei and Elon Musk believe that the pace of creating AI models should be slowed down so that safety rules can be made. At the same time, stalwarts like Nvidia's Jensen Huang and Meta's Mark Zuckerberg are against slowing down the race.