Kuwait Press Memory Latest news
alraiTrending By | إعداد عبدالعليم الحجار|

Which AI platforms are the safest for your data?

Which AI platforms are the safest for your data?

Tom’s Guide published a comprehensive investigative report conducted by one of its editors specializing in digital security and privacy. The editor systematically compared the privacy policies, data collection settings, and data retention practices of all major AI-powered chatbots. To do so, he created new accounts on each platform and examined every available setting and every clause in the lengthy privacy policies that average users rarely read.

The report detailed how a tech expert delved into the settings of ChatGPT, Gemini, Claude, and Microsoft’s Copilot, as well as Perplexity, highlighting three key questions:

• How long is data retained after deletion?

• What options are actually available to users to control their privacy?

The investigation yielded notably divergent results. Some platforms place user privacy at the core of their design, while others bury privacy settings in hard-to-access submenus and enable data collection options by default, which benefit their model development.

Below is a summary of the evaluation for each platform, as reported:

• Anthropic’s Claude platform received the highest rating for privacy and security. It does not automatically use conversation data to train its models and provides a clear interface for fully deleting conversations. Its data retention policy is limited to 90 days solely for abuse monitoring purposes, after which data is permanently deleted.

• OpenAI’s chatbot allows users to disable the use of their conversations for model training via the “Data Controls” menu. However, this option is not enabled by default, meaning any new user contributes their data to training unless they manually take steps to opt out—a fact unknown to many users.

• Gemini ranked relatively low in privacy assessments. Google retains user conversations for up to 18 months, even after turning off activity history. This data is used to improve its models after user identifiers are removed. However, the ambiguity surrounding the de-identification process raises genuine concerns about the potential for re-identifying the data later.

• Copilot is closely tied to the user’s general Microsoft account, meaning conversation data may be linked to the user’s full digital footprint across the company’s various services. This is particularly concerning for users who use the same account for both work and personal purposes.

• Perplexity offers an option to disable data usage for training through account settings. However, the investigation noted that its privacy policy grants the company relatively broad rights over aggregated data, and some AI-powered search features collect usage data that may not be obvious to users.

The report emphasized a crucial point: all platforms, without exception, allow users to manually delete their conversation history. The fundamental differences lie in three factors: default options upon account creation, the duration data remains on the company’s servers after deletion, and the possibility of using deleted conversations to train future models after user anonymization.

The report further noted additional practices worthy of attention, such as some platforms employing human contractors to review samples of conversations for quality assurance purposes. This may expose sensitive content to human eyes, although most platforms now offer an opt-out option for this human review.

The report advised users to take five immediate, practical steps: enable maximum privacy settings immediately upon creating an account; verify that the option to use data for training models is turned off; periodically delete chat history; avoid sharing any sensitive personal, financial, or medical information with any chatbot; and treat all platforms on the assumption that anything typed may be read.

The report concluded that the AI privacy landscape remains fluid and in constant flux, with a disturbingly wide gap between platforms that are most privacy-respecting and those that treat user data as a free commodity. It emphasized that only a conscious user who dedicates a few minutes to reading and adjusting settings can truly maintain control over their data in this new digital space.

Latest news Original source
Link copied ✓