What did Meredith Whittaker say about AI chatbots?

In an interview with Bloomberg, Signal's president stated: 'These are not your friends. These are not conscious beings. These are not sentient interlocutors.' Whittaker warned about the risk of relying emotionally or informationally on commercial systems that collect data.

Why does Whittaker criticize Microsoft Copilot?

Whittaker criticized the vision of Microsoft's AI CEO Mustafa Suleyman, who had described a future in which Copilot manages holiday shopping by accessing family chats, credit cards, browsers, calendars, and Signal. Whittaker called this scenario 'a system with pervasive access across applications and services — which, in Signal, would constitute a backdoor.'

Does Whittaker use AI?

Yes, but in a limited way. She said she uses AI tools 'to format a document here and there,' but doesn't ask chatbots questions and keeps her own thinking process separate from AI. 'I'm very serious about my thinking and my writing, and I don't want the process of working through an idea to be obscured by the response of a system that mediates what already exists.'

Are AI agents really a privacy problem?

The risk Whittaker identifies is real: AI agents operating across multiple applications require broad access — calendar, email, messages, browser, bank accounts. Every integration expands the data-collection surface. The question isn't whether agents work, but who controls the data they collect and how it's used.

Signal and Privacy: Whittaker's Warning on AI Chatbots

Meredith Whittaker isn't the most obvious voice to comment on AI. As president of Signal — the world's most secure messaging app — her main field is privacy, not AI development. But that's exactly why she's worth listening to.

In an interview with Bloomberg, Whittaker said what many think but few say with this much clarity: AI chatbots aren't your friends.

Her Exact Words

Asked about the privacy implications of chatbots like ChatGPT and Claude, Whittaker replied:

"These are not your friends. These are not conscious beings. These are not sentient interlocutors."

It's not a criticism of AI as such — it's an invitation not to confuse a commercial technology with a human relationship. AI chatbots are products built by companies with commercial interests. Their interface is designed to feel conversational, empathetic, helpful. But behind it is a system that collects data, learns from interactions, and operates within business models that aren't necessarily aligned with the user's interests.

The Sharpest Point: AI Agents as a Backdoor

Whittaker's most specific criticism targets the vision of AI agents Microsoft has been promoting. The company's AI CEO, Mustafa Suleyman, had publicly described a future in which Copilot manages users' holiday shopping by listening in on their family chat groups.

Whittaker broke down what this vision would technically mean:

"What you just described is a system with pervasive access across multiple applications and services. In the context of Signal, that would constitute a backdoor."

The point is precise: an AI agent that manages holiday shopping by listening to your family chat groups needs access to Signal, your browser, your credit card, your calendar, your home address. Every single item on this list is a data-collection vector.

The private conversation that Signal encrypts end-to-end — the very thing the app is famous for — would become accessible the moment an AI agent gets permission to "read Signal" to help you with shopping.

Whittaker's Personal Use of AI

Whittaker isn't an anti-AI purist. She said she uses AI tools "to format a document here and there." But she adds:

"I don't ask chatbots questions. I'm very serious about my thinking and my writing, and I don't want the process of working through an idea to be obscured or overshadowed by the response of a system that mediates what already exists."

It's an interesting distinction: using AI as a tool for specific, bounded tasks is different from entrusting it with your own cognitive process or your private conversations.

Why This Perspective Matters

Whittaker isn't selling an alternative product — Signal has no AI agents, no data monetization, no business model based on collecting user information. Her position isn't motivated by commercial competition with OpenAI or Microsoft.

It's the voice of someone who built their career around digital privacy, looking at the current moment and seeing a systemic risk in the normalization of pervasive AI access.

It doesn't mean AI agents are necessarily bad — it means the conversation about privacy, data access, and user control needs to happen before mass adoption, not after.

When in doubt, the question to ask before connecting an AI agent to any account is simple: who manages the data I'm granting access to, how is it used, and what happens if I change my mind?