Is ChatGPT safe?

ChatGPT privacy concerns

Businesses look to use the latest AI technology to improve efficiency and stay competitive. They don't want to be left behind.

OpenAI's ChatGPT is the most widely used AI chatbot known for its advanced features and AI capabilities. Yet, many businesses hesitate to enter business or personal data into ChatGPT because of privacy concerns:

• Does ChatGPT store your data?
• Does ChatGPT share your data?
• Does ChatGPT train on your data?
• Is ChatGPT encrypted?

This article explores these concerns and examines AI services designed for security and privacy.

Does ChatGPT store your data?

Yes, it does. OpenAI uses Microsoft Azure to process and store data, saving conversations on their servers. OpenAI claims to remove deleted or unsaved chats within 30 days. Starting in February 2025, OpenAI will let Enterprise customers store their data in European data centers, making it easier to follow regulations like GDPR.‍

Does ChatGPT share your data?

‍OpenAI is not in the business of selling user data. Their FAQ and privacy policy state that user content isn't shared for marketing or advertising purposes. However, beyond indirect data sharing with their infrastructure provider, they may also share data with:

• Affiliates: OpenAI may disclose personal information to related entities. A complete list of affiliates is not publicly available and may change.
• Law enforcement and government agencies: OpenAI may share data when legally required.


So, while ChatGPT does not share data for marketing, it is not entirely clear who may access user information.

Does ChatGPT train on your data?

‍AI models improve by learning from data. If they are trained on user data, they may retain that information, potentially allowing third-party users to access it later. This raises concerns about whether ChatGPT learns from user inputs.
OpenAI claims its Enterprise GPT products won't train on user data unless explicitly opted in. For ChatGPT users with Free, Plus, and Pro subscriptions, training on user data is the default. However, users have the option to opt out.

‍Is ChatGPT encrypted?

‍OpenAI uses encryption to protect data at different stages:


• At rest: ChatGPT secures data stored on service infrastructure with industry-standard AES-256 encryption.
• In transit: Data moving between a user’s device and OpenAI's servers is protected using modern TLS 1.2+ encryption.


However, OpenAI does not provide details on how they manage encryption keys or who can access them.

‍ChatGPT security claims

‍OpenAI is committed to taking extensive security measures against unauthorized access to user data. Their team passed a SOC2 Type 2 audit and runs a bounty program to encourage finding security bugs. SOC 2 (Service Organization Control 2) is a security framework that assesses how well a service provider protects customer data.
That said, the ChatGPT privacy policy leaves some questions unanswered. It's unclear who concretely has access to chat data and how access might change over time.

‍ChatGPT privacy – can you trust it?

‍The answer to "Is ChatGPT safe for confidential information?" comes down to trust. You must trust OpenAI to protect your data strictly and follow their privacy and security policies once you enter messages into ChatGPT. You have no way of verifying:
‍

• That encryption is correctly applied to stored data
• That your data is genuinely excluded from training
• Against unauthorized access by OpenAI staff or infrastructure providers.


‍While OpenAI is not in the business of selling data, it is in the business of providing knowledge and intelligence. And both get improved by training. The more data they use for training, the better their models become. This creates a strong incentive to collect as much information as possible and incorporate it into training.


Given these uncertainties, businesses handling sensitive information remain cautious about ChatGPT privacy and security. They usually see local AI setups as their only secure alternative.

‍Using local AI setups to avoid security risks

‍Some businesses have started running AI models on their own servers. On-premises solutions improve data protection and privacy while reducing compliance risks (e.g., GDPR, HIPAA). But they also come with challenges:


• High infrastructure costs: Require expensive hardware (GPUs, TPUs) and ongoing maintenance.
• Complex deployment & management: Need expertise in model training, orchestration (e.g., Kubernetes), and optimization.
• Scalability challenges: Harder to scale dynamically compared to cloud-based solutions.


Many businesses are not able to deal with these hurdles. Be it because of financial constraints or a lack of in-house expertise. What they need is an easy-to-use AI cloud service with unquestionable privacy and data protection.

‍Privatemode: The confidential AI with no infrastructure costs

‍Privatemode offers data confidentiality without the infrastructure costs of a local AI. It offers an AI chat that puts user privacy first. It provides similar AI capabilities to ChatGPT but with security that matches on-premises solutions. Privatemode applies Confidential computing, a novel technology that provides security assurances rooted in specialized hardware. Users no longer need to trust the service or infrastructure provider.

How Privatemode ensures confidentiality

• End-to-end encryption: Privatemode encrypts your data at every stage, from submission to AI processing and response.
• No trust in service or infrastructure providers: Privatemode runs in isolated Confidential Computing Environments (CCE) – secure enclaves on external servers that act like locked vaults. CCEs ensure that neither the service provider nor the infrastructure host can access user data.
• Never learns from user data: Privatemode does not use customer inputs for AI training.
• Verifying Service and Data Processing: The app verifies all data protection measures before connecting to the AI service. Hardware-based and provider-independent security and integrity checks ensure a secure and transparent service.
• Transparency: Privatemode is open source, and its security and data protection are auditable for everyone.
• Local hosting: All services are hosted in top-tier data centers within the EU.

Embrace AI without privacy and security risks

With Privatemode, you finally have a confidential AI Chat or API that enables you to:


• Process personal data
• Process proprietary business data
• Meet privacy and data protection regulations.


All without the hassle of setting up and maintaining on-premises solutions. And you still get access to the latest AI models.


You can try Privatemode for free for 14 days. Privatemode also offers developers an API with the same security and privacy features.