Privacy in the Age of AI: Navigating the Complex Landscape of GPT-4o and Data Security

To look at this article again, go to My Profile and then click on View saved stories.

Is It Safe to Share Sensitive Information with GPT-4o?

OpenAI introduced its latest GPT-4o AI technology on May 13, enhancing the ChatGPT service. This updated model boasts impressive skills and a more natural interaction style, including solving mathematical problems, crafting tales for children, and reportedly recognizing emotional states through facial analysis.

OpenAI has expressed a strong desire to provide its tools at no cost to all users. However, professionals argue that the enhanced features of GPT-4o significantly increase the scope of data OpenAI can collect, raising alarms over privacy issues.

The company's history in this field is inconsistent. Following its debut in 2020, a research document demonstrated the use of vast quantities of data gathered from Reddit discussions, literary works, and various internet sources to develop the AI text generation system, incorporating the personal information individuals post online. Consequently, ChatGPT encountered issues with data privacy authorities in Italy, culminating in a short-lived prohibition within the nation the previous year.

Shortly following the introduction of GPT-4o, an early demonstration of the macOS ChatGPT desktop application suggested the possibility of the chatbot accessing a user's screen. Then, in July, concerns arose once more regarding the app when it was discovered that a significant security flaw allowed for the easy retrieval and reading of chat logs stored on a computer in an unencrypted format.

OpenAI promptly released an update to secure chat conversations, highlighting the intense focus on the organization and its latest AI model, GPT-4. This situation raises questions about the privacy of the most recent ChatGPT version. Are privacy concerns greater than they have been with past versions? Moreover, what measures can be taken to enhance its security?

OpenAI's Confidentiality Agreement

At first glance, OpenAI's confidentiality agreement reveals a significant level of data gathering, encompassing personal details, information on how the services are used, and any content submitted during usage. ChatGPT utilizes the information you provide to refine its algorithms, but this can be disabled in the settings or avoided by opting for the enterprise edition.

OpenAI emphasizes in its privacy policy that user data is made "anonymous," yet according to Angus Allan, a senior product manager at the digital consultancy CreateFuture, which provides guidance on utilizing AI and data analytics to companies, their overall strategy appears to be more of a "gather everything first and figure it out afterward" approach. "Their privacy policy clearly indicates they gather all user interactions and maintain the right to use this data to train their models," he notes.

Allan points out that the term "user content" probably includes photos and audio recordings as well. He describes it as an extremely aggressive data collection tool, with all the details explicitly outlined. Although the policy remains largely unchanged in the transition to GPT-4o, the definition of 'user content' has vastly expanded due to its enhanced features.

According to the privacy guidelines provided by OpenAI, ChatGPT is designed to only interact with data that users intentionally share during conversations. Nonetheless, Jules Love, the founder of Spark—a consultancy that guides businesses in incorporating AI solutions like ChatGPT into their operations with a focus on data privacy—points out that ChatGPT inherently gathers a wide array of user information. This includes, but is not limited to, the content of conversations, email IDs, contact numbers, geographical locations, online behaviors, and details about the user's device.

Open AI indicates that the data is employed to refine the AI model and enhance its answers. However, the conditions permit the company to distribute your personal details with partners, suppliers, service entities, and law enforcement agencies. "Therefore, it's challenging to predict the final destination of your data," Love remarks.

According to Bharath Thota, a data scientist and the chief solutions officer of analytics practice at Kearney, a management consulting firm that guides companies on leveraging AI data for new sources of income, OpenAI's privacy documentation reveals that ChatGPT gathers data necessary for account creation or business interactions.

He mentions that the gathered data encompasses complete names, login details, credit card data, and records of transactions. "Individual details could also be retained, especially when users upload pictures in response to certain requests. Similarly, when someone chooses to interact with the company's social media profiles on platforms such as Facebook, LinkedIn, or Instagram, their personal data might be captured if they have made their contact information public."

OpenAI, much like other major technology and social media firms, utilizes user data. However, it diverges from the norm by not engaging in advertising sales. Rather, its focus is on offering tools, a distinction highlighted by Jeff Schwartzentruber, a leading machine learning scientist at eSentire, a security company. He notes that the data provided by users isn't exploited as a direct commercial asset. Instead, it's leveraged to enhance the services provided to users, which, in turn, amplifies the worth of OpenAI's intellectual property.

Data Protection Measures

Following its debut in 2020, in response to backlash and concerns over privacy issues, OpenAI has unveiled measures and settings aimed at securing your personal information. OpenAI asserts its dedication to safeguarding individuals' privacy.

OpenAI acknowledges that some ChatGPT users might prefer their data not be utilized for enhancing its algorithms. As a result, the company offers options for users to oversee their own data usage. According to OpenAI's website, both ChatGPT Free and Plus members have the capability to adjust their settings to opt out of having their data used for further model advancements. Additionally, the company specifies that by default, it does not use data from API, ChatGPT Enterprise, and ChatGPT Team customers for training purposes.

OpenAI offers ChatGPT users various privacy settings, such as a straightforward option to exclude their data from AI model training and a temporary chat feature that ensures conversations are regularly erased, explains OpenAI's representative Taya Christianson to WIRED.

The company states that it does not actively gather personal data for the purpose of refining its algorithms, nor does it utilize publicly available online information for creating user profiles, directing advertisements towards individuals, or for targeting purposes. Moreover, it does not engage in the sale of user information.

OpenAI specifies that it does not use audio from voice conversations to train models unless users opt to contribute their audio for the purpose of enhancing voice chats for all, as stated in the Voice Chat FAQ on OpenAI's website.

"Open AI mentions in its Voice Chats FAQ that by sharing your audio, it could be utilized to enhance their models through your voice chats. Additionally, based on the options you select and your subscription, your conversation transcriptions might also be employed for model training purposes."

In the past few years, OpenAI has improved its openness regarding the gathering and utilization of data "to some extent," offering individuals straightforward choices to adjust their privacy preferences, according to Rob Cobley, a business associate at Harper James, a legal firm that provides counsel on issues of data privacy. "Individuals have the ability to retrieve, revise, and remove their private data, guaranteeing they maintain authority over their information."

According to Love, the simplest method to maintain the privacy of your information involves navigating to your individual settings and disabling the data gathering feature.

Allan advises that it's in the best interest of "nearly all individuals" to take the time to opt out of model training at their earliest convenience. "Opting out won't delete your content from the platform, but it will prevent it from being utilized in the training of upcoming models, which poses a potential risk of your data being exposed."

To accomplish this, navigate to Settings, then Data Controls, and disable the option "Improve the model for everyone."

Additionally, to prevent OpenAI from gathering your data, opt for a "Temporary Chat" for every session. Simply select ChatGPT at the upper left corner and activate Temporary Chat, which you'll find towards the end of the menu.

Nonetheless, restricting the gathering of information can diminish the system's capabilities. "It won't retain any details from past conversations, resulting in responses that are usually broader and lack depth," Love explains.

According to Schwartzentruber, through the ChatGPT online platform, individuals have the capability to erase their conversation records, implement customized directives to enhance privacy protection, manage shared hyperlinks, initiate a data export request, and terminate their account. "To bolster security further, users have the option to enable two-factor authentication and the feature to sign out from all devices simultaneously."

To protect your privacy while utilizing ChatGPT, it's important to be cautious of other aspects. One such aspect includes the inadvertent exposure of confidential information when engaging with Custom GPTs.

Schwartzentruber notes that controlling the information you provide to ChatGPT-4o initially can help you govern the data it collects during interactions. However, he highlights a dilemma between preserving privacy and enhancing user experience. Limiting the flow of data to ChatGPT could impact crucial elements of your engagement with the AI, leading to decreased customization and diminished precision and pertinence in responses, since the AI will rely on a narrower spectrum of basic algorithms.

Latest Update as of July 31, 2024, 8:40 AM Eastern Time: This article has received an update for the purpose of making it clear that the AI model known as GPT-4o serves as the foundation for ChatGPT.

Suggested For You…

Delivered to your email: A selection of our top stories, curated daily just for you.

A faulty update from CrowdStrike brought global computer systems to a halt

Headline News: The Imminent Threat of the Atlantic Ocean's Disruption

Introducing the age of excessive online consumption

Olympics: Stay updated with our comprehensive coverage from Paris this summer right here.

Additional Content from WIRED

Evaluations and Instructions

© 2024 Condé Nast. All rights reserved. Purchases made through our site may result in WIRED receiving a share of the sale as part of our Affiliate Partnerships with retail partners. The content on this website is protected and cannot be copied, shared, broadcast, stored, or utilized in any form without explicit written consent from Condé Nast. Ad Choices

Choose a global website