#ai-privacy

#ai-privacy

If you're at all concerned about privacy, the rise of AI personal assistants can feel alarming. It's difficult to use one without sharing personal information, which is retained by the model's parent company. With OpenAI already testing advertising, it's easy to imagine the same data collection that fuels Facebook and Google creeping into your chatbot conversations. A new project, launched in December by Signal co-founder Moxie Marlinspike, is showing what a privacy-conscious AI service might look like.

Perhaps sensing that feeding more data into Gemini would give many people the creeps, Google's announcement explains at great length how the company has approached privacy in Personal Intelligence. Google isn't getting any new information about you-your photos, email, and search behaviors are already stored on Google's servers, so "you don't have to send sensitive data elsewhere to start personalizing your experience."

"After careful evaluation, we determined that Google's technology provides the most capable foundation for Apple Foundation Models and we're excited about the innovative new experiences it will unlock for our users," Apple and Google said in a statement. The partnership confirms previous reporting on a deal with Google. Neither Apple nor Google have confirmed the price tag, but previous reports indicate Apple could be paying Google around$1 billion for access to its AI technology.

Sometimes, a false sense of intimacy with AI can lead people to share information online that they never would otherwise. AI companies may haveemployees who work on improving the privacy aspects of their models, but it's not advisable to share credit card details, Social Security numbers, your home address, personal medical history, or other personally identifiable information with AI chatbots.

Katharine Jarmul challenged five common AI security and privacy myths in her keynote at InfoQ Dev Summit Munich 2025: that guardrails will protect us, better model performance improves security, risk taxonomies solve problems, one-time red teaming suffices, and the next model version will fix current issues. Jarmul argued that current approaches to AI safety rely too heavily on technical solutions while ignoring fundamental risks, calling for interdisciplinary collaboration and continuous testing rather than one-time fixes.

The report shows that 91% of organizations believe sensitive data should be allowed in AI model training. At the same time, 78% report high concern about theft or breaches. Experts say this discrepancy stems from a lack of understanding about the permanence of data in AI systems. Once sensitive information is used to train a model, it cannot be fully removed or made completely secure. This creates a lasting exposure risk, particularly when personal or confidential data is involved.

Last month, AI startup Friend launched an eyebrow-raising advertising campaign in the New York City subway, which drew a striking amount of hatred. The largely white billboards left a convenient amount of room for passersby to air their feelings about the privacy-infringing tech. As such, it didn't take long for handwritten scribbles to cover the ads. "Befriend something alive," one pen-wielding tagger wrote. "AI wouldn't care if you lived or died," another vandal raged.

Make a playlist with some Latin artists that are on my heavy rotation,

"Users think they're getting a free VPN or SEO widget; in reality, their most private queries - health scares, finances, identity crises - are being slurped, anonymized, and resold," Dryburgh explained in an email. " Onavo and Jumpshot déjà vu, only worse: this time it's your inner dialogue." "We have access to 150+ million real user conversations. This is primarily clickstream data where a user has opted in to be tracked and automatically shared their ChatGPT conversations."

As AI takes off, the whole cycle promises to repeat itself again, and while AI might seem relatively cheap now, it might not always be so. Foundational AI model-as-a-service companies charge for insights by the token, and they're doing it at a loss. The profits will have to come eventually, whether that's direct from your pocket, or from your data, you might be interested in other ways to get the benefits of AI without being beholden to a corporation.

Unique links are created when Grok users press a button to share a transcript of their conversation - but as well as sharing the chat with the intended recipient, the button also appears to have made the chats searchable online. A Google search on Thursday revealed it had indexed nearly 300,000 Grok conversations. It has led one expert to describe AI chatbots as a "privacy disaster in progress".

Le Chat is the least privacy invasive LLM platform, closely followed by ChatGPT and Grok, highlighting differences in data transparency and user opts for model training.