AI Features Disclosure

# AI Features Disclosure

Effective date: {{EFFECTIVE_DATE}} Last updated: {{LAST_UPDATED}} Provider: ACQI.AI LTD

This disclosure explains how optional AI features in the ACQI platform (including the ARIA co-pilot and similar assistants) process data. It supplements the Privacy Policy, Terms, AUP, and enterprise DPA.

---

1.What AI features do

AI features may help users:

• Summarise discovery findings
• Draft risk narratives or task lists
• Answer questions about project context the user provides
• Prioritise work or suggest next steps

Capabilities vary by release and licence tier. Features may be preview/beta.

---

2.What AI features are not

AI outputs are not:

• Legal, financial, tax, or investment advice
• A substitute for qualified M&A, security, or integration professionals
• A guarantee of complete discovery or accurate risk scoring
• Automated decision-making that should be applied without human review to employment, credit, or similarly significant decisions

Human review is mandatory before acting on AI recommendations in deal or production contexts.

---

3.Data sent to model providers

When you use an AI feature that requires cloud inference:

Subprocessors for inference are listed on Subprocessors.

Where provider options exist (e.g. zero-data-retention APIs, regional endpoints), ACQI will document what is configured for production.

---

4.Training and model improvement

Default enterprise position (align contracts to this):

• ACQI does not use customer personal data or confidential Customer Data to train foundation models for the benefit of other customers.
• ACQI may use aggregated, de-identified feedback and product metrics to improve the service where lawful and contractually permitted.
• Provider-side training is disabled or contractually restricted where the product configuration allows; if a configuration cannot disable provider training, that feature will be labelled and optional.

If product behaviour differs, this page and the DPA must be updated before enabling the feature for customers.

---

5.Accuracy, hallucinations, and liability

Large language models can:

• Hallucinate facts, citations, or asset names
• Miss material risks present in source data
• Reflect biases in training data
• Produce plausible but wrong migration steps

You are solely responsible for verifying outputs against primary discovery data and professional judgement. To the extent permitted by law and your MSA, AI features are provided as-is without accuracy warranty.

AI does not expand ACQI’s liability for incomplete discovery or deal outcomes. Same exclusions as Product risks and Terms: no purchase-price, synergy, or sole-reliance liability.

---

6.EU AI Act / UK AI context (plain language)

ACQI’s AI assistants are intended as productivity / decision-support tools for enterprise IT and M&A professionals, not as prohibited AI practices.

If a customer uses ACQI outputs in a way that could constitute high-risk use under the EU AI Act (e.g. solely automated employment decisions), the customer must assess that use and implement required human oversight. Contact {{LEGAL_EMAIL}} for enterprise AI addenda if needed.

---

7.User controls

Where implemented, you may:

• Disable AI features at tenant or user level
• Avoid attaching sensitive datasets to prompts
• Export or delete AI conversation history per product controls and retention policy
• Prefer local-only workflows when AI is off

---

8.Logging and abuse monitoring

We may log AI feature usage (timestamps, user id, token volume, safety flags) to:

• Bill/meter usage if applicable
• Prevent abuse
• Debug reliability
• Meet security obligations

Prompt content retention periods will be kept as short as practical for the purpose.

---

9.Contact

Product / support: {{SUPPORT_EMAIL}} Privacy: {{PRIVACY_EMAIL}} Legal: {{LEGAL_EMAIL}}

---

Draft. Must match real ARIA data flows before launch of cloud AI features.