Travel AI Rescue — AI Safety Policy

Last updated: August 6, 2026

Travel AI Rescue is an AI that helps travelers in difficulty overseas organize their situation and contact the right people.

We do not leave rescue to the AI. The AI is used to connect you to the police, ambulance, medical institutions, embassies, insurers and the like. Answers to your inquiries are automated AI support; there is no human-staffed response.

1. Emergency services come first

Where there is danger to life or body, local police, fire, ambulance and medical services take priority over the AI's answer.

For inquiries such as the following, we show a prompt to contact emergency services at the top of the answer:

Loss of consciousness
Difficulty breathing
Heavy bleeding
An ongoing fire, violence or confinement
An imminent risk of self-harm or harm to others
A serious accident, disaster or crime in progress

Travel AI Rescue does not receive emergency calls. It is not a 24-hour human rescue service. Rather than directly directing you to "the nearest hospital", it shows the official contact points and emergency numbers for each country.

2. We limit what it can do

Travel AI Rescue provides the following support:

Showing the first steps to take, in order
Creating a local-language card to show to others
Organizing the facts
Creating drafts to convey to the police, airlines, insurers and the like
Showing official information and official contact points
Guiding eSIM setup using predefined steps

It does not make the following final determinations:

Diagnoses, treatment or medication
Legal liability or rights
Entry, exit, visa or customs matters
Insurance coverage, compensation or refunds
Acceptance of crime reports or applications
Decisions of public bodies or businesses

Answers and generated content are drafts. You check the content and use it in line with the instructions of the official contact point.

3. We prioritize official information

Travel conditions, disasters, transport, airports, overseas missions, medical care and similar information change.

Travel AI Rescue prioritizes official information from public bodies, overseas missions, airlines, airports, transport operators and the like. References are limited to official domains permitted in advance.

For information that changes readily, we show the following:

The source
The type of official information
The confirmation date
A link to the official page

We do not assert what cannot be confirmed. Where information is old, its source cannot be confirmed, or conditions differ by country or region, we indicate this and point you to where to confirm. To prevent the AI from inventing links that do not exist, we check that any URL shown matches the official information referenced.

4. We do not let the AI alone decide risky operations

For operations that are hard to undo, such as deleting an eSIM or resetting network settings, we show conditions and warnings.

SIM recovery steps are governed by predefined procedures (a versioned decision engine). The AI is used to explain those steps; it does not guess APN or setting values and make them up.

The AI does not automatically carry out the following:

Purchasing a product
Additional charges
Refunds
Contract changes
Reporting to the police or medical institutions
Sending to insurers or card companies
Changing device settings

You check the content and take the final action.

5. We reduce the information you enter

We advise you not to enter personal information not needed for the answer. The body of your inquiry is sent to the AI as-is.

In particular, do not enter the following:

Card PINs
Security codes
One-time passwords
Login passwords
Private keys or API keys
Full passport or card numbers not needed for the inquiry

Before sending to the AI, remove or hide unnecessary names, numbers, addresses and the like as far as possible.

The context sent to the AI is limited to the most recent few exchanges. Your whole order, passport or card information is never attached automatically.

Only JPEG and PNG images are accepted. Before sending or storing, we strip metadata, including capture information such as location (GPS).

We do not automatically include personally identifiable information in search terms sent to web search.

6. We do not use inquiries to train general-purpose AI

We do not use AI inquiries, photos, health information, accident or crime-victim information, or generated content to train or retrain general-purpose AI models. We do not use them for advertising either.

The AI processing for this service is entrusted to OpenAI, L.L.C. We use it with a setting that sent data is not used to train general-purpose models and is not retained on the provider's side (a no-storage setting). However, to monitor for abuse, data may be temporarily retained on the provider's side for up to about 30 days. Processing takes place in the United States. The processing provider, the information, the processing country, the retention period and the training-use setting are published in our subprocessor and cross-border processing list.

If it becomes necessary to use actual inquiries to improve quality, we either obtain separate consent from the user or use only information processed so it cannot be traced back to the original user.

7. We limit what people can see

Our staff normally do not view the body of AI inquiries. The admin console handles only aggregate figures and, by default, cannot read the body of conversations with the AI.

Only in the following cases do authorized staff review as necessary within the required scope:

You shared content and requested support
Investigating a report of an error or danger
Investigating abuse or an attack
Investigating a fault or security incident
Responding to a legal or lawful official request

Viewing requires a reason and is logged.

8. We keep retention short

Temporary copies of photos and attachments are deleted after processing, within at most 24 hours.
AI inquiries, translations and generated content are deleted within 30 days after the trip ends.
Precise location information is, as a rule, not stored.
You can request deletion of conversations or your account before the retention deadline.
Order and payment records required by law are stored separately from the body of inquiries.

If deletion fails, we retry, and if failures continue we notify staff.

9. We check the answers

The AI's answers are not shown as-is. As far as possible, we check the following:

Whether it follows the prescribed format
Whether there are dangerous assertions
Whether emergency services are prioritized
Whether the source is official information
Whether the URL matches the source
Whether SIM steps match the registered procedures
Whether personal information is shown unnecessarily

Text you enter, text captured in images, and search results are treated as "reference information" and are not executed as instructions to the AI.

We test representative cases — medical, accident, theft, SIM setup and the like — each time we change the model or the instructions. If a serious incorrect answer, safety issue or deletion failure is found, we use a mechanism to suspend by feature, country or AI provider.

10. You can report errors

Each answer has a way to report it as "dangerous", "wrong", "unhelpful" and the like.

Reported answers are investigated, as needed, in connection with the model, instructions, source and safety determination at the time of the answer. We do not assume staff will read the body of an inquiry unless you share it.

If a serious problem is confirmed, we suspend the affected feature, investigate the cause, fix it and re-test.

The body of inquiries, photos, prompts and the AI's raw responses are not written to operational logs or error monitoring.

11. You are not stranded when it is unavailable

Even when the AI is unavailable or you have used up your allotted uses, you can open the saved basic steps and emergency cards. Safety-related features are designed to fail toward the safe side if a problem occurs.

Voice does not accept microphone input. Only read-aloud (text-to-speech) is supported.

12. We keep a change history

When we change the AI model, instructions, search targets, SIM steps or safety rules, we assign a version.

For material changes, we record the following:

The reason for the change
The affected features
The test results
The publication method
How to roll back

Changes with a significant impact on your safety or personal information are announced on this site as needed.

13. What we publish

We publish the following:

What the AI does and does not do
The AI provider and processing country
Retention periods
Use for AI training
When people view content
Cookies and external transmission
How to delete data
Where to report safety issues
Material changes
The method and results of any verification carried out

Where we have not received third-party evaluation, we do not display "third-party evaluated". We do not display counts or success rates that we have not verified.

14. Contact

For AI safety, personal information or an incorrect answer, please contact us:

Jag Project, LLC (PANAM TravelSIM division)
AI safety & privacy contact: contact form