OpenAI has taken a meaningful step in the long human negotiation between tool and self, releasing a feature that allows ChatGPT to read through a person's emails and messages in order to write as that person would. Called ChatGPT Work, the capability is available on the web and through paid plans, connecting to platforms like Gmail and Slack to build a model of individual voice and habit. The promise is seamless, personalized assistance; the question it quietly opens is an old one — how much of ourselves we are willing to share in exchange for convenience, and with whom.
ChatGPT Now Learns Your Writing Style From Gmail and Slack
An assistant that already knows your tone means less editing
Mark
So ChatGPT can now read your Gmail and Slack and then write like you do. How does that actually work?
Mimi
It analyzes patterns in your messages—word choice, sentence structure, formality level, recurring phrases. Then when you ask it to write something, it generates text that matches your voice.
Luke
But how much data does it need to build that model? A week of emails? A year? And does OpenAI keep that data, or does it stay on your device?
Mimi
The reporting doesn't specify. We know it connects to Gmail, Slack, and other apps, but the exact scope of what it reads and how long it retains it isn't detailed.
Mark
Why would someone want this? Isn't the point of ChatGPT that it does the thinking for you?
Mimi
It's about speed and consistency. If you're drafting dozens of emails a day, having an assistant that already knows your tone means less editing. You're not retraining it on how you write.
Luke
That's fair, but there's a privacy angle here that feels underexplored. You're giving an AI system access to potentially confidential business communications, personal emails, sensitive information. The feature works because it reads all of that.
Mark
And OpenAI's privacy policy covers it?
Luke
Presumably, but the reporting doesn't dig into what that actually means. What's the retention policy? Can the data be used for training? Is it encrypted end-to-end?
Mimi
Those are the right questions. What we know is the feature exists, it's available on web and paid plans, and it requires connecting your apps. The details about data handling aren't in the public reporting yet.
Le Pouls
- ChatGPT can now read your past emails and Slack messages to learn how you write, then generate new messages that sound authentically like you — without you having to explain your style each time.
- The feature requires granting an AI system access to communications that may contain sensitive, confidential, or deeply personal information, creating real tension around data exposure.
- OpenAI has not clarified how much historical data is analyzed, which apps can be connected, or how the writing-style model it builds is stored or protected.
- The industry is racing toward personalization as the new competitive frontier, and ChatGPT Work is OpenAI's bid to transform the tool from a generic assistant into something that feels like it genuinely knows you.
- The feature is live now for web users and paid subscribers, embedding AI more deeply into the daily rhythm of professional communication than any previous ChatGPT release.
OpenAI has taken a meaningful step in the long human negotiation between tool and self, releasing a feature that allows ChatGPT to read through a person's emails and messages in order to write as that person would. Called ChatGPT Work, the capability is available on the web and through paid plans, connecting to platforms like Gmail and Slack to build a model of individual voice and habit. The promise is seamless, personalized assistance; the question it quietly opens is an old one — how much of ourselves we are willing to share in exchange for convenience, and with whom.
OpenAI has introduced ChatGPT Work, a feature that allows ChatGPT to study your actual writing — by reading your Gmail, Slack, and other connected apps — and then compose messages that sound like you. Rather than adjusting tone on command, the system learns your baseline: the words you favor, your sentence rhythms, how formal or casual you tend to be. It builds a model of your voice, then applies it whenever you ask it to draft something.
This marks a genuine shift in how the tool operates. Until now, ChatGPT could mimic a style you described; now it learns a style you've already lived. The integration with Gmail and Slack is deliberate — these are where professional communication actually happens, and by embedding itself there, OpenAI is positioning ChatGPT less as a separate writing aid and more as an assistant woven into your existing patterns.
The feature is available on the web and through paid subscriptions. But the access it requires is substantial: to learn how you write, the system must read communications that may include confidential business details or personal matters. OpenAI's privacy policies apply, though the company has not specified how much historical data is analyzed or how the resulting voice model is stored.
Behind the feature lies a broader industry logic. As AI models converge in raw capability, the competitive edge increasingly belongs to whoever can make the tool feel most personally attuned. ChatGPT Work is OpenAI's answer — and its wager that users will trade a measure of privacy for an assistant that, at last, sounds like them.
OpenAI has rolled out a new capability within ChatGPT that lets the system study how you actually write—by reading through your Gmail messages, Slack conversations, and other connected applications—and then compose responses that sound like you. The feature, called ChatGPT Work, is available now on the web interface and through paid subscription plans.
The mechanics are straightforward in concept: you grant ChatGPT access to your email and messaging apps. The system analyzes the patterns in your communication—the words you favor, the rhythm of your sentences, how formal or casual you tend to be, whether you use certain phrases repeatedly. It builds a model of your voice. Then, when you ask it to draft an email, a Slack message, or any other piece of writing, it generates text that mirrors not just the content you want to convey but the particular way you would convey it.
This is a departure from how most people have used ChatGPT until now. The model has always been capable of adjusting tone—you can ask it to write something formal or playful or technical. But that adjustment happens in the moment, based on your instruction. ChatGPT Work instead learns your baseline. It doesn't wait for you to specify that you want something casual; it knows you're casual because it has read hundreds of your messages.
The integration touches the tools where most professional communication actually happens. Gmail is where work emails land. Slack is where teams coordinate throughout the day. By connecting to these platforms, OpenAI is positioning ChatGPT not as a separate writing tool but as an embedded assistant that understands the context and cadence of your existing communication patterns.
The feature is available to users on the web and to those with paid ChatGPT subscriptions. OpenAI has not announced restrictions on which apps can be connected or how much historical data the system analyzes to build its model of your writing style.
The capability raises immediate questions about what happens to the data ChatGPT accesses. To learn your writing style, the system must read your emails and messages—some of which may contain sensitive information, confidential business details, or personal matters you would not ordinarily share with a third party. OpenAI's privacy policies govern how this data is handled, but the sheer scope of access required to make the feature work means users are granting the system visibility into communication they might not have considered as feeding into an AI training process.
The move also reflects a broader industry trend toward AI personalization. As large language models become more common, the competitive advantage increasingly lies not in the raw capability of the model but in how well it can be tailored to individual users and workflows. ChatGPT Work is OpenAI's answer to that pressure—a way to make the tool feel less like a generic service and more like an assistant that knows you.
Citations marquantes
OpenAI says ChatGPT can now write like its human user, watching their emails and messages— OpenAI announcement
Envie de l'histoire complète ?
Lire l'original sur Google News ↗
Technology