AI Transparency
Last reviewed: 3 September 2026
PacketSafari combines deterministic packet processing with optional AI-assisted analysis. This page explains where AI is used, what it can and cannot do, and which responsibilities remain with PacketSafari and its customers.
This information is provided for transparency. It is not a certification, conformity assessment, or legal opinion.
Where PacketSafari uses AI
PacketSafari uses third-party general-purpose AI models within an experience designed and controlled by Ripka Technologies S.L. AI-assisted features include:
- Copilot, a conversational assistant for questions about packet captures and derived evidence;
- Agent, an investigation workflow that can select tools, examine bounded evidence, form hypotheses, and produce a report;
- model-authored explanations, findings, summaries, recommendations, and visual reports; and
- API, streaming, command-line, or export formats when they contain those model-authored results.
At the start of a Copilot or Agent surface, PacketSafari identifies Copilot and Agent as AI systems, warns that output may be incorrect, and asks users to verify important conclusions against cited packet evidence. The notice is intentionally compact and remains visible in the relevant workflow.
What is not AI-generated
PacketSafari's Core Engine and Triage functions perform deterministic packet decoding, extraction, indexing, counting, rule evaluation, and bounded correlation. Exact packet fields, frame references, filters, protocol statistics, and deterministic findings are not model-authored merely because an AI feature later uses them as evidence.
We keep that distinction visible because a generated explanation and an extracted packet fact have different reliability and verification requirements.
Intended use
PacketSafari is intended for network troubleshooting, packet analysis, and cybersecurity investigation. It is not intended to make decisions about a person's employment, education, credit, access to essential services, biometric identity or traits, law-enforcement treatment, migration status, healthcare, or another regulated individual decision.
Do not use PacketSafari for unlawful surveillance, manipulation, social scoring, biometric categorisation, predictive policing, or another prohibited AI practice. A proposed new purpose near a prohibited or high-risk boundary requires review before release or use.
Limits and human oversight
AI output is probabilistic. It can be incomplete, inaccurate, inconsistent, or overconfident, even when it cites real packet evidence.
Users should:
- verify important conclusions against cited frames, filters, deterministic findings, capture scope, and timestamps;
- confirm that the available capture contains the traffic needed to support the conclusion;
- keep a qualified human responsible for operational, security, legal, and safety decisions; and
- apply their organisation's access, retention, confidentiality, incident-response, and change-control requirements.
PacketSafari is decision support for qualified analysts. It does not replace professional judgement or required monitoring, detection, validation, or safety controls.
Deployment responsibilities
| Deployment | PacketSafari | Customer |
|---|---|---|
| Managed SaaS | Operates the integrated product and approved model route; controls the application experience and bounded evidence supplied to AI features. | Controls authorised users, uploaded captures, purposes, prompts, decisions, and organisational oversight. |
| Teams SaaS | Operates the shared managed team service and approved hosted model route. | Controls organisation access, use, customer policy, and decisions based on output. |
| Dedicated SaaS | Provides a customer-specific managed environment and the model/data route stated in the applicable agreement. | Controls enterprise access, approved purpose, customer policy, and decisions based on output. |
| On-premises | Supplies the integrated software, product instructions, limitations, and release updates. | Operates the deployment, configures approved AI endpoints and egress, controls data and retention, and manages authorised users. |
Ripka Technologies S.L. is likely the provider of the integrated PacketSafari AI system and a downstream provider when third-party general-purpose models are incorporated. A customer using PacketSafari in its organisation is generally a deployer. Exact legal roles can depend on the contract, deployment, rebranding, or a material modification.
AI-generated output and marking
Visible AI labels help people understand the origin and limitations of an output. Machine-readable marking is a separate provider obligation for covered synthetic audio, image, video, and text output, subject to the applicable legal exceptions.
PacketSafari is implementing a consistent marking contract for covered model-authored text, reports, images, API responses, streams, command-line output, and exports. Existing visible labels or report-integrity metadata should not be interpreted as proof that every output format already has the machine-readable marking required by Article 50(2).
Deterministic packet extracts and facts are not synthetic output. Assistive editing or output that does not substantially alter supplied input may fall within a legal exception, but exceptions are assessed by output class rather than assumed for the product as a whole.
Data and privacy
Packet captures can contain personal data, credentials, payload content, and confidential network information. The data path varies by deployment and configuration. Review the Privacy Policy, Subprocessor List, Trust Center, and the applicable customer agreement before enabling cloud-hosted AI for sensitive data.
Customer documents and updates
The Terms of Service govern service use. Deployment guides and API documentation describe release-specific controls and data paths. Contractual terms, privacy statements, and technical documentation must match the actual model route and deployment selected by the customer.
We review this page when PacketSafari's AI features, model routing, intended uses, output-marking approach, or relevant official guidance materially changes.
Official sources
This page is informed by current official primary material, including:
- Regulation (EU) 2024/1689, consolidated text;
- Regulation (EU) 2026/1744;
- European Commission guidance on Article 50 transparency obligations;
- European Commission AI literacy questions and answers; and
- European Commission guidance on prohibited AI practices.
Questions about PacketSafari's AI-assisted features can be sent to contact@packetsafari.com.