OpenAI's latest move against Anthropic: Promise not to retain customer data OpenAI promises not to retain corporate customer data using its artificial intelligence model, and at the same time improves product security capabilities, aiming to steal customers from competitor Anthropic

OpenAI’s latest move against Anthropic: Promise not to retain customer data OpenAI promises not to retain corporate customer data using its artificial intelligence model, and at the same time improves product security capabilities, aiming to steal customers from competitor Anthropic. Many companies are dissatisfied with the developer of Claude’s practice of saving their own business data. ChatGPT developer said on Wednesday that it has opened a preview of the new technology to some customers. The technology can identify behavioral patterns and security risks across multiple rounds of model interactions. In the past, AI security systems could only monitor individual interaction records one by one. Without retaining some customer data, it would be difficult to ensure that tools are used compliantly; the new solution aims to break through this limitation. This commitment is in sharp contrast to Anthropic’s current policy: the latter will retain customer data for 30 days to ensure the safety of the latest model operations. This approach has been criticized by many technology industry leaders, including White House AI advisor and venture capitalist David Sachs, Microsoft CEO Satya Nadella, and Palantir’s Alex Karp. They proposed that enterprises need to obtain deterministic guarantees to confirm that AI manufacturers will not save their own data. This demand is particularly prominent in highly regulated industries such as medical care and finance. Alia Hauser, head of product policy at OpenAI, said: "The demands of enterprise customers are very clear, and this matter is crucial to them. Many companies themselves have to bear data responsibility for their own customers; these service providers who serve other companies have a large amount of highly sensitive data." This move reflects OpenAI's desire to catch up with Anthropic in the AI ​​​​race, while responding to criticism. Critics worry that the company's products are not adequately protected against cyberattacks or the various hazards caused by out-of-control models. According to reports, OpenAI recently revealed to investors that its second-quarter revenue growth lagged behind Anthropic, disappointing some shareholders. OpenAI said that customer data will be retained in the infrastructure controlled by the customer. The company is also developing another system: data can be stored by OpenAI, but the encryption key belongs to the customer. This means that OpenAI employees cannot view user prompt words, model responses, and other sensitive information. When the system identifies high-risk activities such as cyber attacks and biological weapons related activities from the model usage behavior, it will send a signal to OpenAI to determine whether disposal actions need to be taken. Customers can appeal the results of the disposition and provide more information to OpenAI as needed. OpenAI only saves this highly summarized signal record, marking only the risk level and risk category. via cnBeta.COM - Chinese industry information station (a