ChatGPT Adds Trusted Contact, Fast Answers, and Memory Sources

OpenAI just shipped the biggest ChatGPT consumer update since GPT-5.5 launched. The release notes posted this week include three changes that meaningfully shift how the product behaves: a Trusted Contact safety feature for mental-health crises, a "Fast answers" low-latency mode, and editable memory sources that let users see and change what ChatGPT remembers about them.

Taken together, the updates point at where OpenAI is steering ChatGPT for the rest of 2026 — toward more personalization, more transparency about that personalization, and more guardrails for the moments when a chatbot is the wrong tool for the job.

Trusted Contact: The Safety Move Critics Have Demanded

Trusted Contact is the most consequential addition. Users can now designate a person — a family member, a friend, a partner — who can be notified by OpenAI in "serious suicide-related safety concerns." The feature is opt-in, available only on personal accounts in supported regions, and explicitly excluded from Business, Enterprise, and Edu workspaces.

The rollout follows two years of public pressure from mental-health researchers, families of users in crisis, and regulators in the EU and California. Chatbots have been at the center of multiple wrongful-death lawsuits and policy debates over how AI should respond when a user is in distress.

OpenAI's framing is deliberately narrow. Trusted Contact does not turn ChatGPT into a crisis service; it gives ChatGPT a way to escalate a serious moment to someone who actually knows the user. Why this matters: it shifts the responsibility for crisis response from an AI making judgment calls to a human who has context — exactly what clinicians have been asking for.

Fast Answers: A Speed Tier for Common Questions

The second change is more pedestrian but probably affects more users daily. Fast Answers is a new mode optimized for "common information-seeking questions with high-confidence, in-depth replies." It works globally on web, iOS, and Android, skips past chats and memory, and can be turned off in Personalization settings.

In practice, Fast Answers is OpenAI catching up to Perplexity and Google's AI Mode on the latency war for search-style queries. When you ask "what's the capital of Mongolia" or "how do I unclog a drain," you don't need ChatGPT to consult your memory and chat history before answering. Fast Answers strips out that overhead.

The toggle matters. Power users who want their assistant to know them will leave Fast Answers off. People using ChatGPT as a search alternative will likely leave it on. OpenAI is acknowledging — for the first time, really — that those are two different products.

Memory Sources: You Can Finally See Why ChatGPT Said That

The third update is the one privacy researchers have been asking for since Memory shipped in 2024. ChatGPT now shows the actual sources behind each response: which past chats, saved memories, files, and connected Gmail messages shaped the answer.

You can click into the source list, see exactly what context ChatGPT pulled in, and edit or delete it. That last part is the new bit. Previously, users could see that Memory existed but had limited ability to surgically remove what they didn't want there.

The feature also expands the memory surface itself. Plus and Pro users now get personalization that pulls from past chats, saved memories, files, and connected Gmail accounts. Context-aware responses are richer; the transparency layer is what makes the richness palatable.

Industry Impact: Personalization With Receipts

The combined effect is a clear shift in OpenAI's positioning. ChatGPT is no longer just a chatbot — it's a personal AI that knows you, with explicit controls for what it knows and an escalation path when something goes seriously wrong.

That positioning puts pressure on competitors in three directions. Google's Gemini Spark is the obvious agent-side rival, but it doesn't yet have an equivalent of Trusted Contact. Anthropic's Claude has stronger enterprise governance (this week's Compliance API launch) but weaker consumer personalization. Meta's AI inside Instagram and WhatsApp has reach but minimal transparency about memory.

For users, the practical result is that ChatGPT becomes harder to switch away from — your personalization travels with the account — while also becoming easier to audit. That's a deliberate trade.

Expert Perspectives

Mental-health professionals welcomed Trusted Contact as a concrete step beyond the generic "if you're in crisis, call this hotline" responses chatbots have defaulted to. Several noted that the opt-in design is critical: forcing escalation would erode user trust, but offering it gives users in vulnerable moments a real fallback.

Privacy advocates were more cautious. Editable memory sources are a real improvement, but the underlying question — how much of an AI assistant's personalization is reversible, and how much gets baked into model behavior — remains unanswered.

What's Next

The Codex preview in the ChatGPT mobile app, also rolled out this week, suggests OpenAI is unifying its consumer and developer surfaces. Expect Trusted Contact to expand to more regions in coming months and to potentially extend to other crisis categories beyond suicide-related signals.

You can enable Trusted Contact in the ChatGPT mobile or web app under Settings → Personalization. Memory sources are visible inline when ChatGPT responds. Fast Answers shows up as a toggle for new chats.

Bottom Line

OpenAI just made ChatGPT meaningfully more personal, meaningfully more transparent about that personalization, and meaningfully safer at the worst moments. It's the clearest sign yet that the company has moved past "ship features, sort it out later" and is starting to act like a steward of how a billion-plus people use AI every day. If you've been on the fence about turning on Memory, this is the update that makes it worth a try.