SAIL, Safe AI at Lancaster, is Lancaster University's AI service for staff and postgraduate researchers. It is operated by the ISS Innovation team and provides access to approved AI models for teaching, research and professional services work.

Your access is linked to approved use cases. These describe what you can use SAIL for, what information you can enter and what checks you need to make.

No. SAIL is an ongoing ISS service with no fixed pilot end date. Your access remains linked to your approved use cases, and the service will be reviewed and maintained as it develops.

SAIL provides a University-managed environment for AI use. It supports safe experimentation and approved operational or research use cases, while reducing risks associated with unapproved external tools or personal AI subscriptions.

SAIL is intended to help the University manage access, data conditions, assurance, human review and accountability consistently.

SAIL is available to Lancaster University staff and postgraduate researchers whose access has been approved. Each account is linked to one or more approved use cases.

Postgraduate researchers must submit a use case and receive approval before an account is created. Access requests are approved by the Head of Innovation and Mobile Development.

A use case describes what you want to use SAIL for, who will use it, what information is involved and how the results will be checked.

Once approved, it sets out what you can do and any conditions you must follow. If your purpose or the information you use changes, request a review before proceeding.

A use-case owner is the person accountable for the purpose, operation and periodic review of an approved use case. They may work with technical, data protection, security, research governance or other specialist colleagues where required.

Yes, where the activity falls within an approved use case and you have the appropriate access.

SAIL may support tasks such as:

• Drafting, editing and improving documents
• Summarising and analysing approved material
• Planning projects, meetings or research activities
• Exploring ideas and alternative approaches
• Interrogating approved knowledge bases and sources
• Developing reusable prompts, templates, skills or workflows
• Producing outputs such as summaries, reports, analyses or code

You remain responsible for checking outputs and complying with University policies, local procedures, data protection obligations, research ethics requirements and any contractual or funder conditions.

SAIL is intended to support productivity, innovation and responsible experimentation. It does not replace professional expertise, academic judgement, decision-making or accountability.

Human review remains essential before relying on AI outputs or taking actions that may affect people, services, research, records or decisions.

Submit an https://www.lancaster.ac.uk/itpi/sail/ describing what you want to use SAIL for. Access is granted, once the request has been approved and linked to an approved use case. Postgraduate researchers must submit their own use case for approval.

Complete the required introductory briefing before using SAIL.

Yes. You can access SAIL from any location using a supported web browser and your University credentials, subject to the normal University security requirements.

Yes. SAIL is browser based and available from any location, provided you sign in with your university credentials.

Your account is your identity within SAIL. It is linked to your University credentials and to any approved access, roles and permissions.

A role is a defined set of responsibilities or permissions within SAIL. Examples may include user, editor, owner or administrator.

Permissions determine what actions you may take on a resource. For example, you may be permitted to view, use, edit, share, approve or administer a workspace, workflow or knowledge base.

Yes. Access is based on approved roles and permissions. This may affect which use cases, workspaces, models, tools, knowledge bases, integrations or sharing options are available to you.

A resource being visible in SAIL does not automatically mean that every user has permission to access or use it.

Where your University account has multiple roles, Your access is determined by the permissions approved for your account and use cases. Having more than one University role does not automatically give you access to additional resources. Request additional access if your work requires it.

Where access is not already available, you may need to submit an access request for SAIL or a specific use case. Access will be granted only after the appropriate approval has been recorded.

https://www.lancaster.ac.uk/itpi/sail/

A workspace is a purpose-specific environment for approved work. It may bring together related conversations, files, sources, instructions, workflows and outputs within the scope of a use case.

Workspaces may have their own access rules, data conditions and ownership arrangements.

You can upload files such as Word documents, PowerPoints, or PDFs and ask questions about its content. This is helpful for summarising, analysing, checking structure, or explaining material.

You may upload personal or sensitive information if it is part of your normal work. SAIL will highlight if it detects personal or sensitive data to help you stay aware and make an informed choice. These alerts do not block your work.

Uploaded content is stored securely for your session and deleted automatically after a short retention period.

A project keeps related work together. It can contain chats, workflows and associated files for a piece of work. Use a named project to organise work you want to return to.

A project does not change the permitted purpose or data conditions of your approved use case.

A chat, or conversation, is an interaction with SAIL within an approved use-case scope. It consists of messages from you, the AI assistant and, where relevant, other authorised components.

A prompt is an instruction or request supplied to guide the AI’s response or work. A prompt may be written by an individual user, provided through a template, or built into an approved skill or workflow.

A task is a defined piece of work with a goal, scope, constraints and expected result. For example, a task might be to produce a plain-English summary of an approved policy document.

Where available, SAIL history and search features enable you to find and return to previous work that you are authorised to access. Availability will depend on the data-retention arrangements, your permissions and the relevant use case.

The assistant is the AI experience that supports users through conversation, explanation, drafting, analysis and refinement.

The assistant can help with approved tasks, but it does not replace subject expertise or professional judgement.

An agent is an AI component configured to pursue a delegated goal through a series of steps and, where permitted, by using approved tools.

Agents may be available only in particular approved workflows or use cases. Their actions, permissions and required human approval points must be clearly defined.

In SAIL, a skill is a reusable AI task component. It provides guidance or capability for completing a particular type of work consistently and may include instructions, expected inputs, constraints and output requirements.

In this context, “skill” does not refer to a user’s qualifications or level of proficiency.

A model is the underlying AI system used to interpret requests and generate responses. SAIL provides access only to models that have been assessed and approved through University governance arrangements.

Models are selected following assessment of security, data protection, governance, licensing, capability, accessibility, cost and responsible AI considerations.

Model versions and configurations may also be tested and approved before use. Where appropriate, model information may be displayed to support transparency.

A model router is the internal mechanism that selects an approved model for a task according to configured requirements and constraints. In some cases, you may be able to select from approved models yourself.

Tools perform defined operations to support a task. Examples may include extracting text from a document, searching approved sources, calculating values or transforming content.

Tools may only be available within approved use cases or workflows.

Instructions define expected behaviour, process, boundaries, style or output requirements for an AI interaction, skill or workflow.

Templates are reusable structures for prompts or outputs, such as a report format, briefing template or review form.

SAIL may provide a transparency panel or equivalent information about the capabilities, limitations, availability, cost or sustainability implications of available models. This is a proposed service improvement and may not yet be available in all areas.

You may be able to upload supported file types, including:

• Microsoft Word documents (.docx)
• Microsoft Excel workbooks (.xlsx)
• Microsoft PowerPoint presentations (.pptx)
• PDF documents

You can then ask SAIL to summarise, explain, compare or analyse the content, where this is permitted by the relevant use case and data conditions.

Before uploading a file, check that you are authorised to use it and that it does not contain information that must not be entered into SAIL.

Conversation context is the messages and material available within the current conversation. It helps SAIL respond to the task you are working on.

Workspace context is the files, sources and instructions associated with a workspace. It may be available to authorised users working within that workspace.

A source is a document, dataset, web page or other reference used to support a task or answer. Sources may be supplied by users, stored in a knowledge base or accessed through an approved integration.

A knowledge base is a managed collection of approved sources. It has defined ownership, access rules, data conditions and review arrangements.

Knowledge bases are intended to support more consistent responses based on approved information. You should still refer to the original source where a decision, policy interpretation or formal action depends on the information.

A knowledge-base owner is the person or team responsible for ensuring that a knowledge base remains suitable, accurate, maintained and appropriately accessible.

Retrieval is the process of finding relevant information from authorised sources to support a task or answer. SAIL may show the retrieved evidence or citations where this is available and appropriate.

Check the original source before relying on a citation, quotation or reference. AI can provide inaccurate or invented references, and a real source may not support the claim being made. For decisions or policy interpretations, confirm the information in the authoritative source.

Memory refers to information retained across conversations to inform later interactions. Persistent memory is not assumed to be available in SAIL and, if introduced, will be subject to defined governance, data conditions and user information.

SAIL is configured so that prompts and responses are not used by approved external AI providers to train their models. The specific arrangements for each component are subject to the University’s contractual, technical and governance controls.

SAIL stores conversations and related session data securely for a limited period to support use of the service. Session data is automatically deleted after 30 days, and you may be able to delete your own conversations earlier.

For full details, refer to the SAIL Privacy Notice:

[SAIL Privacy Notice Link Pending]

Your conversations are not visible to other SAIL users unless you explicitly share them or work within an authorised shared workspace.

Access to service data is managed in accordance with University security, operational and legal requirements.

The University does not routinely read individual conversations. However, limited system, security and audit information may be processed where necessary to operate, secure, investigate misuse of or improve the service, in line with the SAIL Privacy Notice and relevant University policies.

An assurance record contains evidence of relevant specialist review, such as security, data governance, accessibility, research governance or data protection assessment outcomes, including Data Protection Impact Assessment (DPIA) outcomes where applicable.

Assurance records are primarily managed internally. Relevant conditions and restrictions will be communicated to users of the relevant use case.

Data conditions are the rules that define what information may be used in a use case, how it may be handled and what is excluded. They may include restrictions relating to personal data, confidential information, research data, commercially sensitive information or third-party material.

Personal or sensitive information may be used only where that use has been reviewed with Data Governance and approved for your use case. Follow any conditions set through that review.

Access to SAIL does not, by itself, give permission to enter any type of information. If you are unsure whether your information is covered, check before uploading it.

Yes. You must comply with:

• The approved use-case scope and data conditions
• University Responsible Use of AI guidance
• The University Acceptable Use Policy
• Data protection and information-security requirements
• Research ethics, governance, funder and contractual requirements
• Any local faculty, department or professional requirements

Before entering or uploading information, consider whether it is:

• Necessary for the task
• Appropriate and authorised for use in SAIL
• Possible to anonymise, redact or minimise
• Subject to confidentiality, contractual, legal or ethical restrictions
• Covered by a data-sharing agreement, ethical approval or Data Management Plan

Responsible Use of AI guidance:

[Responsible AI Use Link Pending]

Personal data is information relating to an identified or identifiable person. It can identify someone directly or indirectly, including when combined with other information. Examples include staff or student records, contact details and research participant information.

Some personal data, including health information and information about ethnicity or religious beliefs, receives additional protection. Follow the data conditions for your approved use case.

The warning helps you check what you are about to submit. It does not give permission to use the information, and it may not detect every instance of personal information.

Before continuing, check that the information is necessary and permitted for your approved use case. Remove unnecessary personal details.

No. AI-generated content may be inaccurate, incomplete, biased or fabricated. This is sometimes referred to as “hallucination”.

You must critically evaluate all outputs before relying on or sharing them. This includes checking facts, figures, quotations, references, legal or regulatory statements and claims about University policies.

No. Safeguards can reduce some risks, but no AI system can eliminate errors, bias or inappropriate outputs. Human review and critical judgement remain essential.

Yes, where the activity is covered by an approved use case. SAIL can support tasks such as exploring ideas, drafting and analysing approved research material, or developing code and analytical approaches.

Use cases involving personal or sensitive information must be reviewed with Data Governance and approved before that information is used.

For guidance on using AI in learning and assessment, including academic integrity requirements, see:

https://portal.lancaster.ac.uk/ask/study/developing-academic-skills/using-ai-in-your-learning-and-assessment/

PGRs completing taught modules or assessments must follow the relevant assessment guidance.

University position on Artificial Intelligence | ASK – Lancaster University:

https://portal.lancaster.ac.uk/intranet/cms/ai-strategy-development/interim-guidance

Where acknowledgement is required, follow the relevant departmental, programme, assessment, publication, professional-body or funder guidance.

Guidance on acknowledging AI use:

[RAG / Acknowledgement Guidance Link Pending]

An integration is an approved connection between SAIL and another University or external system or service.

Integrations are subject to appropriate security, data protection, technical and governance review.

A connection is a configured instance of access to an integrated service. It is subject to authorisation and may be limited to particular use cases, workspaces, roles or users.

SAIL does not automatically access University systems, datasets or records unless there is an approved integration or connection for a specific use case.

You should not assume that SAIL can retrieve, use or act on information in another system unless this has been explicitly enabled and approved.

Connected information is external information made available through an approved connection.

A connected action is an operation performed in another system through an approved integration, such as creating, updating, sending or publishing information.

Connected actions are subject to separate approval, permissions and human review requirements where appropriate.

Credentials and service identities are used to establish authorised system access. They are managed internally and must not be exposed, shared or entered into prompts, chats or documents.

Status describes the lifecycle or execution state of an item. For example, a workflow, use case or output may be marked as draft, approved, running, completed or retired.

An attention flag is separate from status. It indicates that an item may need review, action or intervention.

SAIL may maintain audit history for important access, configuration, change, execution and approval events. This supports service operation, security, accountability and governance.

The extent of user-visible history will depend on the resource and your role.

SAIL may apply quotas or rate limits to manage service capacity, cost and fair access. These may vary by use case, model, tool or user role.

SAIL may collect usage reporting information, such as use of models, tools, workflows, retries, costs and operating activity. Access to this information is controlled according to role and operational need.

Each use case will be reviewed against its intended benefit and the process it supports. Reviews will consider output quality, results accepted for use, time spent including checking and correction, cost and user feedback. This helps decide whether a workflow should continue, change or be extended.

Yes. The Innovation team will maintain SAIL and introduce improvements based on approved needs, user feedback and service reviews. New features will be introduced where they meet governance requirements and can be supported within the service's capacity and budget.

Planned features are not a delivery commitment. Guidance will be updated when changes become available.

Yes. You can request a workflow through SAIL. Explain the task, who would use it, the information involved and the benefit you expect.

The Innovation team will assess its value to the University, feasibility and any governance requirements. Submitting a request does not approve the workflow or guarantee that it will be developed.

You can submit feedback and requests through SAIL. You can also add ideas through the University Ideas Wall:

https://www.lancaster.ac.uk/itpi/web/services/ideasWall/topics/247https://www.lancaster.ac.uk/itpi/web/services/ideasWall/topics/252

Requests are considered through the appropriate technical, security, data protection and governance processes.

For technical problems, contact the ISS Service Desk through the University's usual support channels. Describe the problem and when it happened. Do not include sensitive conversation content in a support request unless you are asked to provide it through an approved route.

For questions about permitted use or information, follow the contact route provided for your approved use case.

SAIL operates within Lancaster University’s data protection framework. Your rights under UK data protection law may apply to personal data processed through the service.

These rights can include:

• The right to be informed about how your data is used
• The right to request access to your personal data
• The right to request correction of inaccurate data
• The right to request erasure in certain circumstances
• The right to request restriction of processing in certain circumstances
• The right to object to certain processing
• The right to complain to the University or the Information Commissioner’s Office

You may be able to manage some information directly, for example by deleting conversations where this option is available. Session data is also deleted automatically in line with the retention period described in the SAIL Privacy Notice.

For further information, consult the SAIL Privacy Notice:

[SAIL Privacy Notice Pending]

If you have concerns about how your personal data is handled, contact the University Data Protection Officer through the official University channels.