AI policy for external consultants

Date: November 2025 Applies to: All external contractors, freelancers, and consultants working with E Co.

1. Policy Statement

E Co. is committed to fostering a productive and responsible working environment. We recognise that Artificial Intelligence (AI) has the potential to increase efficiency, improve decision-making, and drive innovation. However, we also recognise that these benefits come with risks regarding data protection, content quality, copyright, and confidentiality.

As a freelance consultant working with E Co., you are an extension of our values. This policy outlines the principles and strict guidelines you must adhere to when using AI technologies in the delivery of services to E Co.

The purpose of this policy is to:

Promote the secure and ethical use of AI technologies in work delivered to E Co.
Ensure compliance with relevant laws and regulations.
Safeguard E Co.’s proprietary data, as well as the privacy of our employees, customers, vendors, and beneficiaries.
Establish clear accountability regarding the ownership and quality of AI-generated work.

2. Application of this Policy

This policy applies to all external consultants, freelancers, and contractors (collectively "Consultants") engaged by E Co.

Compliance with this policy is a mandatory requirement of your engagement with E Co. Failure to adhere to these guidelines, particularly regarding data privacy and confidentiality, may be considered a material breach of contract and could result in the immediate termination of your service agreement and potential legal action for damages.

3. Core Principles

Your use of AI in relation to E Co. projects must adhere to the following principles:

A. Ethical Use: AI must be used in a manner that respects human rights and diversity. You must ensure fair and unbiased outcomes.
B. Data Privacy and Security: strict adherence to data protection laws and E Co.’s confidentiality requirements is non-negotiable. E Co. Confidential Information must never be exposed to public AI models that use user data for training.
C. Transparency: You should be transparent about the use of AI in your deliverables where appropriate.
D. Human Oversight (Human-in-the-loop): Human expertise must remain the driving force of your work. AI is a tool to enhance your capabilities, not a substitute for your professional judgement.
E. Accountability: You remain fully responsible and liable for the accuracy, originality, and legality of all work submitted to E Co., regardless of whether AI tools were used in its creation.

4. Acceptable Use and Tool Selection

Approved Applications

E Co. acknowledges that freelancers often utilise their own suite of tools. However, when processing E Co. Confidential Information, strict restrictions apply:

Public Tools (e.g., standard ChatGPT, or other free AI models): You are strictly prohibited from inputting E Co. Confidential Information, Personally Identifiable Information (PII), or client data into public/free AI tools where data is used for model training.
Enterprise/Private Mode: If you use AI tools to process E Co. data, you must ensure the tool is operating in a mode where data retention for training purposes is disabled (e.g., Gemini Pro, MS Copilot Pro, ChatGPT Enterprise).
E Co. Provided Tools: If E Co. provides you with access to internal AI tools, you must use them strictly in accordance with the training and guidelines provided.

If you are unsure whether a specific tool is safe for E Co. data, you must consult your E Co. point of contact before use.

Functional Assistance

AI may be used for drafting, ideation, and planning. However, the final deliverable must be reviewed, verified, and approved by you. You must act as the "Human in the Loop" at all times.

5. Managing Risks

Data Privacy and Leakage

The Risk: Public AI tools often "learn" from user input. If you paste a confidential E Co. project brief into a chatbot to generate a summary, that brief may become part of the AI's public knowledge base.
The Rule: Assume all data input into an AI tool is non-confidential unless you have verified otherwise. Never input names, addresses, trade secrets, or specific client details into an unsecured AI tool. Use generic placeholders (e.g., "Client X", "Project Y") if you must use these tools for reasoning.

Copyright and Intellectual Property

E Co. requires full ownership of the IP for work commissioned.

Originality: You must ensure that AI-generated content does not infringe on third-party copyrights.
Verification: Large Language Models (LLMs) can hallucinate or reproduce copyrighted text verbatim. You must cross-reference and fact-check all AI output.
Liability: As the Consultant, you indemnify E Co. against any claims of copyright infringement arising from AI content included in your deliverables.

Bias

AI models can reflect historical biases. You are responsible for carefully evaluating all AI-generated content to ensure it is free from discriminatory language or bias regarding age, disability, gender, race, religion, or sexual orientation.

6. Training and Competency

As an external expert, E Co. expects you to be competent in the tools you use. This includes understanding the privacy settings of your AI software. However, E Co. may provide specific guidance regarding our internal data classification. If you are unsure about E Co.'s specific data handling requirements, please pause and ask your E Co. contact immediately.

7. Review and Contact

This policy is subject to periodic review. If you have questions regarding the permissibility of a specific tool or use case, please approach the technical contact as listed in your contract.

Glossary

Confidential Information: Information not publicly known, including sensitive business info, client data, and trade secrets.
Chat Training: A feature in AI tools where the model learns from conversations. This poses a high data leakage risk.
Human in the Loop: The requirement for human judgement to verify AI output.
PII (Personally Identifiable Information): Data that can identify an individual (names, emails, etc.).

Report abuse