AI Usage Policy

Artificial intelligence (AI) is evolving rapidly, and its fast adoption means it will be used in all aspects of human life, including communications. Continuous breakthroughs in AI model development have already transformed how individuals and organizations across industries generate content, analyze data, and create value for stakeholders.

The practice of augmenting decision-making, copy review, and visual mock-up development, among others, is now commonplace in the communication industry.

Clients have become increasingly aware of AI role in business workflows, including in communications consultancies. They have begun requesting clear disclosures of AI use and made it binding in legal agreements.

To remain a trusted, competitive, and forward-thinking partner, Vero is introducing its AI Usage Policy—a structured framework that defines how AI can be used ethically and responsibly across the organization. This policy sets out Vero’s approach to AI, including its risk inventory, governance practices, and commitment to safeguarding human creativity and integrity in all communications work.

This policy aims to provide guidance for AI use within Vero’s work environment. The goal is to have a productive interaction with AI tools that enhances work quality, efficiency, and competitiveness while ensuring responsibility over sensitive information, professional integrity, and client trust.

It is developed by following available best practices, applicable regulations, and industry conventions such as the ICCO Warsaw Principles. This will allow Vero to ensure full compliance with legal standards and client expectations.

This policy applies universally to all Vero team members, including full-time staff, contractors, freelancers, interns, and consultants. Regardless of role or seniority, every team member is expected to understand and comply with these guidelines whenever engaging with AI as part of their daily responsibilities.

The policy encompasses any interaction that AI is used for, but not limited to, internal communication, client service, content development, analytics, research, or third-party collaboration.

Literatures (Ng, D. T. K., 2021; Gil de Zúñiga et al., 2023) tend to define AI as a replication of intelligent behavior by machines. This behavior includes understanding language, making calculations, predictions, interpretation, plans, or creating art.

Due to its wide coverage, this policy will deal mostly with the use of AI tools – that is software made to perform intelligent tasks. Examples of such tools are discussed below. Any tools or software not mentioned in the following section but perform similar behaviors will also be classified as AI tools, and their use is also subject to this policy.

The following risk assessment is driven by an internal survey augmented by secondary data sources.

Used AI Tools

These are the known AI tools used within the Vero work environment, by order of appearance frequency.

1. ChatGPT
2. Canva
3. Gemini
4. Copilot
5. Grammarly
6. Claude
7. Apple Intelligence
8. Quillbot

Known AI Uses

These are among the reported uses of AI in the  Vero work environment, by order of appearance frequency.

1. Checking grammar and translation
2. Drafting email correspondence
3. Research
4. Excel formula
5. Brainstorming ideas
6. Copywriting
7. Editing articles
8. Management advisory
9. Legal advisory
10. Tax advisory

Known AI Input

Aside from prompts, the following body of information is often put into AI tools to produce information.

1. Marketing plans
2. Original drafts
3. Key messages
4. Document templates
5. Audience persona
6. Trend reports
7. Media reports
8. Published marketing material

Known AI Risks

The following risks are put together based on the survey responses and categorized using academic concepts.

1. Platform-bound risks

a. AI hallucination (fabricated information): Even the latest AI systems often generate plausible-sounding but entirely fictional statements, known as hallucinations, which may be mistaken for real facts if not carefully vetted. Failure to anticipate this risk will result in the spread of fabricated quotes or inaccurate data that can undermine credibility and trust.

• Example: A consultant asks ChatGPT for references to back a client’s insight, and the AI returns non-existent article citations that appear legitimate.
• Mitigation: Two-step human review and requirement of corroborating evidence.

b. Biased outputs: AI systems rely on their training data set, and due to various factors existing globally, these data sets have been identified to be biased. For example, data sets in English are much more available than those of ASEAN local languages, thus the perspective of English speakers is much more represented in the AI’s process than others, resulting in biased results.

• Example: An AI assistant drafting key messaging for a campaign about public health awareness in Southeast Asia frames the content through a Western-centric lens, omitting culturally relevant values and local idioms, resulting in poor resonance with the target audience.
• Mitigation: Two-step human review and requirement of local culture expertise.

c. Legal/IP risks due to unclear AI training data: Some generative models may be trained on copyrighted material, raising potential infringement issues when their outputs are used commercially.

• Example: An image generated by AI for a campaign includes stylistic elements from a copyrighted brand logo, resulting in a cease-and-desist letter.
• Mitigation: Seek AI tools that are trained on proprietary datasets (e.g., purchased or owned stock visual assets) instead of problematic sources (e.g., copyrighted films, social media users’ photos).

2. User-bound risks

a. Disclosure of sensitive or confidential data: Inputting proprietary or private client information into AI systems may risk exposure or breach of confidentiality agreements because it can be used for purposes other than what was intended.

• Example: A team member enters confidential product details into generative AI tools, which risks violation of the client’s NDA and leaks confidential content because the data could be used to train future AI models.
• Mitigation: Use of secure working environment with no training guarantees and adherence to minimum required information principle.

b. Over-reliance on automated content without human validation: Dependence on AI outputs without appropriate human oversight may lead to misaligned messaging, off-brand tone, or strategic misjudgments. Furthermore, as AI tools tend to produce what is significant in their data set, results that represent the middle of the bell curve tend to appear more often, resulting in similarities.

• Example: A press release draft produced by AI is sent to a client without edits and includes phrasing that contradicts the client’s public stance or substance that is completely out of context.
• Mitigation: Two-step human review.

c. Shadow AI (unauthorized tool usage): Vero team members may use unvetted tools outside company policy, creating hidden compliance and data security risks.

• Example: A team member uses a free AI tool found online to draft content, unknowingly exposing confidential brand research.
• Mitigation: Provide access to approved tools and device-based blocking if possible.

d. Transparency failure and erosion of trust: Clients or partners may discover unacknowledged use of AI in their deliverables, potentially undermining trust in Vero’s professional standards.

• Example: A client identifies AI-generated content in a published report and raises concern that no disclosure was provided.
• Mitigation: Early disclosure and proactive client education.

Risks Perception Rating

The following is AI risk perception among Vero team members.

The risk’s likelihood is measured on a 0-10 scale, with 0 being “doesn’t happen” and 10 being “happens all the time”. Likewise, the risk’s impact is measured on a 0-10 scale, with 0 being “no impact” and 10 being “disastrous”.

It is important to note that Vero team members all choose the higher part of the scale when assessing the risks; thus, the above chart only shows the first quadrant (impact and likelihood from 5-10).

The perception of AI risks shows that hallucination is viewed with the highest awareness, despite having only moderate likelihood and impact scores, making it the most visible concern in daily use.

Legal/IP/Plagiarism and privacy breaches stand out as the most severe threats, with privacy breaches carrying the highest potential impact, while legal/IP issues are seen as both likely and highly consequential.

In contrast, risks such as bias, over-reliance, shadow AI, and trust erosion are clustered at moderate levels, acknowledged but not prioritized. Overall, the focus lies in balancing safeguards against high-severity risks while addressing the more visible but less catastrophic challenge of hallucination.

Privacy & Data Protection

Team members are mandated to only use recognized AI tools that have clearly stated their commitment to privacy and data protection. The use of public, anonymous, or anonymized information in interactions with AI is highly recommended. The use of identifiable information must adhere to the principle of minimum required information.

Quality & Accuracy

Vero uses AI systems alongside older, proven methods (e.g., academic repositories, industry reports, stakeholder engagements) to ensure working teams have sufficient breadth and depth to produce great recommendations. This wealth of knowledge is cross-verified, compiled, and synthesized by a team of human beings.

Ethical Use & Bias Mitigation

Vero does not tolerate deceptive, manipulative, or misleading behavior with or without AI tools. AI outputs are always reviewed by multiple team members with sensitivities to local cultures and nuances, who are also subject to the company’s code of ethics and professional conduct.

Human Oversight

Vero’s position is that AI tools are used to support—not replace—professional judgment. All AI outputs are reviewed and validated by Vero team members, who remain fully responsible for the final deliverable.

Transparency & Disclosure

Vero will disclose all material use of AI proactively and clearly through contract clauses with clients or AI disclosure statements embedded in relevant documents.

The main criteria for approval of AI tools are data security. Vero does not approve the use of AI tools that use its input and output for purposes other than the users’ intention, including but not limited to, AI training dataset and marketing purposes.

Approved

Approved after Opt-Out

Out-out guides are included in the attachment.

Vero will disclose AI usage early and proactively. Disclosure statements will be embedded in contract clauses, and in the absence of an active contract, disclosure statements will be embedded in relevant documents. The disclosure statement will be clear on how AI may contribute to the respective work and how the work as a whole complies with Vero AI Usage Policy.

Please see the attached disclosure template for more details.

Permitted Uses

Internal

• Summarizing meeting notes or briefs: Transcribe and distill key takeaways from internal discussions, interviews, and stakeholder calls.
• Generating initial drafts: Produce early versions of memos, reports, or emails, creating a structured starting point. All content must be reviewed and refined by humans before delivery.
• Workflow automation within Office apps: Team members may use Microsoft 365 Copilot for generating presentation outlines, compiling repetitive data entries, or transforming tables into charts, saving time on repetitive formatting.
• Brainstorming content ideas: Support ideation for campaign soundbites, headlines, or creative angles during internal planning sessions. The final concept must be reviewed, edited, and have its responsibility owned by the respective team lead.
• Research assistance: Scan media coverage, identify patterns in public discourse, or extract high-level insights from large reports. Human teams must validate findings manually and contextualize them for strategic relevance.
• Sentiment tracking and media monitoring: Categorizing brand sentiment or scanning media for specific themes or anomalies. Internal analysts validate the results to inform communications strategy.
• Crisis signal detection: Detect emerging issues or public backlash early through abnormal media spikes or sentiment changes, triggering internal escalation protocols.

Client-Facing

• First drafts of deliverables: Generate structural drafts that are later customized by teams with creative directions, substances, and facts, while ensuring alignment with brand voice, legal requirements, and news value.
• Supporting visualizations and content structure: Suggest frameworks for infographics, reports, mock-ups, or slide decks, which human teams adapt into client-ready assets.
• Trend analysis: Support data scanning to identify media trends or sentiment patterns, which are then interpreted and refined by consultants to form actionable insights.
• Presentation scripting: Outline talking points for client presentations or internal reviews, with final delivery tailored by human speakers.
• Data clustering and thematic analysis: Group large sets of articles, posts, or survey responses into insight-driven clusters. Consultants then define strategic implications based on business objectives.
• Media landscape scans: AI can help compile relevant outlets, influencers, or publication trends to support PR strategy development. Final recommendations are refined by communications experts.

Prohibited Uses

Internal

• Inputting confidential or sensitive data: Team members must be careful in handling confidential information with AI tools. Use only the minimum necessary information to complete the task and it must be done in secure working conditions.
• Using unvetted or unauthorized AI tools: Only tools approved by the AI Committee may be used for internal tasks. AI platforms that lack privacy guarantees are not permitted.
• Relying on AI without human review: Any internal content that is reused externally or escalated to clients must pass through a human validation process. Raw AI-generated text must never be used as-is.

Client-Facing

• Impersonation or identity simulation: AI must not be used to mimic real individuals (e.g., crafting messages as a CEO or journalist) without explicit consent. This includes recreating writing styles to simulate attribution.
• Misleading or deceptive content: AI must never be used to fabricate data, quotes, or endorsements. All content must reflect truthful information backed by verifiable sources.
• Undisclosed AI-generated material: Deliverables that include AI-generated text, data, or analysis must be transparently disclosed to clients as outlined in the disclosure procedure.
• Culturally insensitive outputs: Content created with AI must be reviewed for local relevance and appropriateness by an expert with close proximity to said locale.

Vero AI Committee

The Vero AI Committee is a cross-functional team responsible for governing AI policy implementation. The committee evaluates and approves new tools, reviews usage trends, updates risk controls, and ensures staff are trained in safe, effective AI practices. It meets quarterly or more frequently if needed to assess evolving needs, regulatory changes, or incident response.

The members are:

1. Agnes Alpuerto
2. Brian Griffin
3. Fuad Arrasyid
4. Nguyen Thi Thao Chi
5. Thanpitcha Suksawad
6. Market representatives: Indonesia, Philippines, Singapore, Thailand, Vietnam

Incident Reporting Procedure

• Team members who saw any problem with their AI interaction should develop an incident log and inform the committee within three business days.
• Incidents may include, but not be limited to, data leaks, offensive/biased output, use of shadow AI, misuse of sanctioned AI tools, misinformation, and bias.
• Reports can be submitted anonymously to committee members in each market or to any member of the AI Committee. Once escalated, the AI Committee must convene to decide on the incident.
• Information that should be included in the incident log: description of the incident, AI tools involved, and any preliminary remediation.
• The committee is responsible for investigating and documenting, as well as supporting the resolution of the incident.

Regular Audit & Review of AI Usage

The AI Committee will conduct surveys to audit AI tools’ effectiveness, policy implementation, and identify any problems on a yearly basis. The results of this survey will inform policy updates, AI tools subscription, and future AI-related training.

Experimental Usage

In addition to tool evaluation, incident review, and training, the AI Committee is tasked with monitoring emerging technologies and governance practices. This includes exploring potential adoption of:

• AI detection tools to help verify originality and authenticity of external content (e.g., influencer submissions, third-party reports).
• AI-supported crisis simulation tools for scenario planning, sentiment modeling, or risk response rehearsal under strict human review.
• Generative audio/video safeguards, ensuring that synthetic media (e.g., voiceovers, avatars, re-creations) are transparently disclosed and reviewed for consent, accuracy, and cultural sensitivity.

Regulatory Alignment

Vero’s AI Usage Policy aligns with and will continue to adapt to national data privacy regulations, including but not limited to:

• Singapore Personal Data Protection Act (PDPA)
• Thailand Personal Data Protection Act (PDPA)
• Indonesia Personal Data Protection Law (PDP)
• Vietnam Cybersecurity Law
• Myanmar Telecommunications Law
• Malaysia Personal Data Protection Act (PDPA)
• European Union General Data Protection Regulation (GDPR)

The AI Committee, with support from the Legal team, will monitor relevant updates and ensure compliance when using AI across regional offices.

Industry Best-Practice Alignment

Vero commits to operating under globally recognized standards for ethical AI and communications, including:

• The ICCO Warsaw Principles for ethical use of AI in PR and communications.
• Guidelines from the Public Relations Council (US) and other international bodies regarding disclosure, fairness, and human oversight.
• Ongoing benchmarking against evolving best practices in transparency, fairness, and accountability in AI-enhanced creative and strategic work.

Violation Handling

Policy violations will be handled in accordance with Vero’s disciplinary framework. Depending on the severity of the infraction, this may include a formal warning, mandatory retraining, or, in serious or repeated cases, termination of employment. Transparency is maintained while balancing due process and privacy.

Upon client request, Vero commits to ensuring no AI tools are used in any phase of a designated project. This includes ideation, drafting, data analysis, and reporting. The compartmentalization happens at team level and does not extend to Vero as a whole.

Compartmentalized teams under “no-AI” agreement will only use traditional tools – such as Microsoft Office, Outlook, and Teams. Generative systems will not be used together with their digital infrastructure.

In principle, vendors contracted by Vero must adhere to the same AI standards and ethical practices defined by Vero AI Usage Policy. The following vendor-facing summary and checklist can be used when discussing AI with a vendor.

Artificial intelligence (AI) is increasingly integrated into communication workflows. Vero recognizes the importance of aligning its external vendors with the same ethical, secure, and transparent AI practices followed internally. This ensures a consistent standard of quality, responsibility, and client trust.

All vendors contracted by Vero—including freelancers, production houses, key opinion leaders, and event organizers—must follow the AI guidelines outlined in Vero AI Usage Policy. If AI is used in any stage of the deliverable process (e.g., drafting, summarizing, editing, image generation, etc.), the vendor must:

• Clearly disclose the tool and its function.
• Ensure full human review of all AI-generated outputs.
• Minimize the use of confidential information with AI tools.
• Use only approved tools (e.g., ChatGPT Enterprise, Microsoft 365 Copilot, etc).
• Notify the Vero team if any risks, bias, or anomalies are observed.

Failure to comply may result in rejected deliverables, contractual review, or discontinuation of collaboration.

Checklist of points that vendors must guarantee:

• Disclosure of all use of AI tools
• Expert review of all AI-generated content, ensuring accuracy, contextual fit, and cultural sensitivity
• No entry of confidential or sensitive identifiable information into AI platforms without privacy guarantees
• Vero-approved AI tools only
• All AI-related incidents will be reported to Vero

The attachments are provided for informational purposes only and should be used responsibly.

Human Review Protocol for AI-Assisted Deliverables

To uphold the integrity and quality of Vero’s client work, all AI-assisted outputs must undergo a structured human review process before being shared externally. This protocol ensures that AI use complements—never replaces—professional judgment.

Step 1: Confirm AI Contribution

• Identify the specific parts of the deliverable influenced or generated by AI (e.g., draft summaries, presentation scripts, visual frameworks).
• Note the AI tool used.

Step 2: Evaluate Materiality

Determine the significance of the AI’s contribution:

• Minor: basic phrasing, spellcheck, formatting.
• Moderate: structured drafts, brainstorming prompts.
• Material: entire sections, insights, or visuals that shape deliverable value.

Step 3: Human Editing Pass

All AI content must be reviewed and edited by the assigned team lead or editor. Key checks include:

• Alignment with brand voice, tone, and client expectations.
• Accuracy and factual correctness of any claims.
• Completeness and logical flow.
• Removal or clarification of any vague, repetitive, or generic AI phrasing.

Step 4: Cultural & Ethical Review

Assign a reviewer with local expertise (for ASEAN markets or global context) to assess:

• Cultural appropriateness and sensitivity.
• Avoidance of stereotypes or implicit bias.
• Inclusivity in language and representation.

Step 5: Final Approval

• Final output must be signed off by the project manager or senior consultant before delivery.
• Check if a disclosure statement is active on the contract level. If not, include a disclosure statement.

Step 6: Incident Logging

Check if in the work process, there was an incident with an AI tool that could be logged and reported to the committee

Templates

• AI Incident Log Template
• Policy Feedback Template
• Vendor Declaration Template
• New Tools Application Template

Opt-Out Guide

As of 2 September 2025

• ChatGPT Free: Your Profile > Settings > Data Controls > Improve the model for everyone > Switch off the toggle.
• Gemini Free: Expand Menu (three-line button at the top left corner) > Settings & Help > Activity > Gemini Apps Activity > Turn off.
• Perplexity Free: Account Settings Page > Preferences > AI Data Retention > Switch off the toggle.
• Canva Free & Enterprise: Homepage > Account Profile > Settings (gear symbol) > Personal Privacy > Switch off the four toggles.

AI Disclosure Usage Boilerplate for Proposals

Vero may use artificial intelligence (AI) tools to enhance the efficiency, quality, and strategic depth of its work. These tools are used strictly under human oversight, in a secure environment, and in compliance with Vero’s internal AI Policy and relevant data protection standards. All AI-assisted contributions are subject to human validation and ownership.

Standard Contract Clause for MSAs

Vero may, where appropriate and beneficial, utilize artificial intelligence (AI) to support the development of content, research, analysis, and related consultancy deliverables. All such AI-assisted work will be conducted in accordance with Vero’s internal AI Usage Policy, which emphasizes data privacy, human oversight, cultural sensitivity, and transparency.

Vero shall ensure that:

1. All data is handled within a secure work environment;
2. Only tools with reputable security and non-training guarantees are used; and
3. All AI-generated outputs are reviewed, edited, and finalized by qualified human professionals.

The Client retains the right to request, in writing, a “no-AI” workflow for designated projects, in which case Vero will assign a non-AI team and work tools, as well as apply strict internal compartmentalization.

References

1. Gil de Zúñiga, H., Goyanes, M., & Durotoye, T. (2023). A Scholarly Definition of Artificial Intelligence (AI): Advancing AI as a Conceptual Framework in Communication Research. Political Communication, 41(2), 317–334. https://doi.org/10.1080/10584609.2023.2290497
2. Ng, D. T. K., Leung, J. K. L., Chu, K. W. S., & Qiao, M. S. (2021). AI literacy: Definition, teaching, evaluation and ethical issues. Proceedings of the association for information science and technology, 58(1), 504-509.