ChatGPT Agent is active on some European accounts, but with no official announcement
- Graziano Stefanelli
- Jul 25
- 4 min read
In recent days, several users in Italy, Germany, Spain, and Switzerland have reported that the ChatGPT Agent mode is now active, despite OpenAI's support pages still stating that it is unavailable in the EEA and Switzerland. This silent rollout has generated confusion and interest across the community, suggesting that OpenAI may be conducting a limited trial before making a broader public release.
OpenAI had explicitly excluded Europe and Switzerland at launch.
On July 17, 2025, OpenAI officially launched ChatGPT Agent, a feature that allows the chatbot to perform complex tasks on third-party websites and apps on behalf of the user. According to the original release notes and continuously updated support documentation, Agent mode was not available in the European Economic Area (EEA) or Switzerland, due to unresolved regulatory constraints.
The help center still explicitly states:
“⚠️ Agent mode is currently not available to users located in the European Economic Area (EEA) or Switzerland.”
The reasons for this exclusion include compliance with the GDPR, the Digital Markets Act (DMA), and especially the upcoming EU AI Act, which imposes strict requirements for autonomous systems capable of performing real-world actions such as purchases, bookings, form submissions, and automated browsing.
Yet over the past 48 hours, dozens of users have reported activation within Europe.
Between July 24 and 25, 2025, numerous users began reporting that Agent mode appeared on their Plus or Pro accounts located within European countries and without using any VPN.
The most frequent confirmations have come from:
Germany: Over 15 users confirmed Agent mode appeared, some for several hours, others briefly.
Italy and Spain: Around 10 users each reported access, many on Android mobile devices.
France, the Netherlands, Austria, Belgium, Poland: Smaller but consistent confirmations.
Switzerland: 3 Pro users confirmed Agent access, though most still report it as unavailable.
In total, over 65 users in Europe have claimed to have Agent mode activated, while about 50 others in the same regions report that it is still unavailable or was briefly activated and then removed.
The geographic distribution is inconsistent: even among users with the same subscription plan and country, some see the feature and others don’t. This strongly suggests a canary rollout or A/B testing, where OpenAI activates and deactivates the feature for select users to gather usage and compliance data.
The legal and technical reasons that delayed the European launch.
Agent mode enables actions with significant real-world implications. For instance, the agent can:
Access websites while simulating user behavior,
Navigate forms, portals, and closed systems,
Execute cross-app operations using stored credentials,
Save data and handle persistent sessions.
This raises serious regulatory concerns in Europe, where platforms must offer:
Explicit legal bases for all data processing (GDPR Article 6),
Auditable logs of automated actions (AI Act and DMA),
Granular opt-in consent mechanisms for any action outside OpenAI’s domain,
Full transparency in interactions with external platforms, which is not yet fully guaranteed.
Independent analysts have noted that OpenAI appears to be adjusting by introducing:
Memory turned off by default,
User permission screens each time the agent accesses a site or service,
Downloadable action logs (currently in testing).
The rollout is not official yet, but might become so soon.
Despite numerous user confirmations, OpenAI has not yet published any formal announcement extending Agent availability to the EEA or Switzerland. The help center page — updated within the last 24 hours — still states that the feature is not available in these regions.
However, the feature is clearly appearing for verified accounts in Europe without VPNs, which
suggests something is changing:
OpenAI is actively testing a release, selectively enabling the feature for a small number of European users.
The rollout could be revoked at any time, or expanded more widely in the coming days.
Users currently seeing Agent mode should treat it as experimental and unsupported for now.
OpenAI’s typical rollout pattern (as seen with DALL·E 3, Memory, and Browse) follows this structure: silent activation, limited test group, and later a formal announcement along with updated support policies.
Realistic short-term scenarios.
Scenario | Likelihood | Detail |
Official announcement of EU availability | ★★★☆☆ | Possible by the end of July if current testing runs smoothly. |
Feature rollback for EU test accounts | ★★☆☆☆ | May happen if compliance issues arise. |
Staggered release: Pro first, then Plus, then Team | ★★★★☆ | A likely strategy already used for other regulated features. |
Availability only in EU countries with lighter enforcement | ★★☆☆☆ | Less likely, but possible as a middle-ground solution. |
How to check if you have access and what to do next.
Check for the “Agent mode” option in the left sidebar or under “Explore GPTs”.
Try creating a custom agent by clicking “Create” and seeing if the “Actions” tab is available.
Look for tags or badges like “EEA Preview” or “Experimental” in the footer — these often indicate early-stage rollout.
Avoid relying on it for business-critical tasks, as the feature might be revoked or disabled without warning.
Do not store sensitive or regulated data until auditability, permissions, and retention are fully documented.
Summary
ChatGPT Agent has quietly appeared on some European accounts in what appears to be a limited internal test. OpenAI is likely preparing for a broader release, potentially following final compliance checks. But as of today, there is no official support or guarantee of availability in the EEA or Switzerland. Users who have access are essentially participating in a live pre-release — and should treat it accordingly.
___________ FOLLOW US FOR MORE.
DATA STUDIOS