“Building the AI model is quite shocking and leaving a backdoor wide open from a security perspective,” says independent security researcher Jerremiah Fowler, who was not involved in visa research, but specializes in discovering exposed databases. “This type of operational data and the ability of anyone who has an internet connection to access it and then manipulate is the main risk for organizations and users.”
Deepseek systems are apparently designed so that they are very similar to OPENAI, scientists said Wired on Wednesday, perhaps to go to up-to-date customers to employ Deepseek without difficulty. They say that the entire Deepseek infrastructure imitates OpenAi to details such as the API key format.
Wiz researchers say they do not know if anyone found an exposed database, but it would not be a surprise, considering how straightforward the discovery was. Fowler, an independent researcher, also notes that the sensitive database “definitely” would be quickly found – if it was not – regardless of other researchers or bad actors.
“I think it is a wave of AI products and services that we will see in the near future and how they seriously treat cybersecurity,” he says.
Deepseek had a global influence over the past week, and millions of people came to the service and pushed her to the top of Apple and Google stores. The resulting shock waves recorded billions of stock prices of American AI companies and terrified management in companies around the country.
On Wednesday, the sources at Openai announced Financial timesThe company analyzed the alleged employ of the company’s CHATGPT results for training Deepseek models. At the same time, Deepseek increasingly draws the attention of legislators and regulatory bodies around the world, who began to ask questions about the company’s privacy policy, the impact of its censorship and whether its Chinese property ensures concerns about national security.
Chinese Deepseek connections also seem to raise, perhaps inevitable security concerns. At the end of last week, according to CNBC reportingThe US Navy issued a notification of its staff, warning it so that they do not employ the deepseek services “to any extent”. E -Mail stated that naval staff members should not download, install or employ the model and raise concerns about “potential safety and ethical”.
However, despite the noise disclosed, the data show that almost all technologies based on databases hosted in the cloud can be sensitive by straightforward safety shortcomings. “AI is a new limit in everything related to cyber technology and security,” says Ohfeld from Wiz, “And still the same old gaps, such as open databases, can still exist open on the Internet.”