ChatGPT: companies veto AI to protect data and secrets – 05/19/2023 – Tech

Apple has banned employees from using ChatGPT and other generative artificial intelligence (AI) platforms, out of concern about sensitive data being leaked. The company is working on its own AI to address the problem, according to a report in the Wall Street Journal this Thursday (18).

Samsung, JPMorgan Chase, Deutsche Bank and Verizon have made similar decisions. Amazon not only vetoes the use of the OpenAI chatbot, but also recommends that its employees use the company’s AI (Code Whisper) to assist with programming tasks.

Other companies such as Accenture say the use of artificial intelligence can violate company governance and give instructions to their employees.

Samsung made the decision to veto the use of ChatGPT and other generative AIs, such as Bing (Microsoft) and Bard (Google), after workers shared company trade secrets with the platform.

An internal document from the Korean electronics manufacturer obtained by Bloomberg indicates that the company feared not having control over the storage of internal information on third-party servers.

For the professor at the Faculty of Law of USP Juliano Maranhão, the data leaked to artificial intelligences cannot even be erased. As ChatGPT prompts, for example, are used to train the AI ​​model, the platform learns the delivered information.

“Artificial intelligence does not do a search like Google does. It is a language processing model that delivers a response from a command. There is no way to delete information”, says Maranhão, who specializes in artificial intelligence.

AI models, however, need a lot of data on a given topic to deliver accurate answers.

Sought by the report, OpenAI did not respond to the request for information until the publication of the text.

The startup provides the option to disable artificial intelligence training while using ChatGPT since the last 25th. This option disables the history of conversations with AI, which limits the experience.

According to the platform’s doubts area, the startup is working on a business-oriented solution to allow turning off AI training without giving up history.

The company also states on its website that it discards data stored by ChatGPT every 30 days. In the chatbot settings, it is possible to download a report of the information kept by the company.

Apple also recommended that its developers not use GitHub Copilot, artificial intelligence that assists in writing computer code.

Sought, GitHub declined to comment on the case. According to the tool’s website, Copilot stores the codes and uses them to train the AI ​​itself only in individual plans, sold at US$ 10. Users of the business package do not deliver information from their programs to the platform.

Maranhão states that companies, in addition to being concerned with disclosing confidential information, need to be aware of cases of plagiarism and disclosure of inaccurate information by those who use generating artificial intelligence. “All this needs to be checked.”

The chief executive of digital marketing company Clearlink, James Clarke, said he was still concerned about remote workers using AI to buy time. Without supervision, they could take on multiple jobs. The statement was part of a leaked internal Clearlink video.

SEE COMPANIES THAT HAVE BANNED THE GENERATOR

Company	Reason
samsung	Disclaimer regarding the leakage of strategic information and trade secrets
apple	Disclaimer regarding the leakage of strategic information and trade secrets
Deutsche Bank	Bank and customer data security
verizon	Bank and customer data security
Northrop Grumman	Disclaimer regarding the leakage of strategic information and trade secrets
amazon	Recommends using the Code Whisper tool itself
JPMorgan Chase	Limitation of use of external software imposed by bank policy
Citigroup and Goldman Sachs	Limitation of use of external software imposed by bank policy
Bank of America	Limitation of use of external software imposed by bank policy