---
title: "Continuous AI agent testing: change tests and quarterly retests"
canonical: https://pharosone.ai/knowledge/continuous-testing
publisher: PharosOne (Agentic Mind, Inc.)
updated: 2026-10-22
license: "Open for reuse: crawling, copying, quoting, summarising and model training permitted without asking; citation of the title and canonical URL requested (see https://pharosone.ai/terms#reuse)"
cite_as: "\"Continuous AI agent testing: change tests and quarterly retests\". PharosOne, 2026-10-22. https://pharosone.ai/knowledge/continuous-testing"
---

# Continuous AI agent testing: change tests and quarterly retests

> From €3,000 a month, billed yearly: the agent changes you hand over are attacked before release, and the whole agent is retested each quarter for AIUC-1.

URL: https://pharosone.ai/knowledge/continuous-testing · Category: Products · Published: 2026-10-22 · Updated: 2026-10-22 · Author: PharosOne Research · Reviewed by: PharosOne editorial

*PharosOne sells continuous testing and wrote this article.*

**In short.** Continuous testing is a yearly subscription under which PharosOne tests the changes you hand over to your AI agent before they ship and retests the whole agent every quarter, with each quarterly report written as AIUC-1 evidence. The volume of change tests and their turnaround are fixed in the SOW. It is for teams that change their agent often enough that a point-in-time pentest is out of date before the next audit. From €3,000 per month, billed yearly.

## Key facts

- Updated: 22 October 2026
- For: teams that ship agent features every sprint and need test evidence that is current when a reviewer asks
- Price: from €3,000 per month, billed yearly, net of VAT; the exact number is fixed in the SOW after a 30-minute scoping call
- Cadence: agent changes tested before release, at the volume and turnaround fixed in the SOW; the whole agent retested every quarter
- You receive: a findings report before each release; a dated quarterly retest report with the attack success rate, the AIUC-1 control mapping and the evidence transcripts
- Standards: AIUC-1 (quarterly retests, annual re-certification); EU AI Act Art. 9; OWASP Agentic Top 10
- Next step: book a 30-minute scoping call at audit@pharosone.ai

## What is continuous testing?

Continuous testing is adversarial testing of an AI agent on a schedule tied to the agent's own changes, as opposed to a single engagement with a start and an end date. An agent changes with every new tool, prompt edit, model upgrade and data source, and each change can open an attack class that the last test never saw. A pentest report dated six months ago describes an agent that no longer exists.

AIUC-1 states a cadence outright: certification includes "quarterly retests on your AIUC-1 Standard Version" and "requires annual re-certification" ([aiuc-1.com](https://www.aiuc-1.com/); [standard.aiuc-1.com](https://standard.aiuc-1.com/)). Article 9 of the EU AI Act describes the risk management system for high-risk AI as "a continuous iterative process planned and run throughout the entire lifecycle of a high-risk AI system, requiring regular systematic review and updating", and requires testing "at any time throughout the development process, and, in any event, prior to their being placed on the market or put into service" ([artificialintelligenceact.eu, Article 9](https://artificialintelligenceact.eu/article/9/)); it applies to high-risk systems and fixes no frequency. ISO/IEC 42001 asks for review through its Plan-Do-Check-Act loop, also without a fixed frequency ([ISO/IEC 42001 in our standards crosswalk](/research/ai-standards-crosswalk#iso-iec-42001)). A quarterly retest meets the AIUC-1 cadence; for the other two, the frequency is yours to justify.

## Who is it for?

- Teams that ship agent features every sprint, where the agent of the last pentest and the agent in production have diverged.
- Teams pursuing AIUC-1 certification, which needs quarterly test evidence on the certified version.
- Teams whose agent falls under the EU AI Act's high-risk obligations and who must show testing before release and systematic review after it.
- Teams whose customers' security questionnaires ask for the date of the last test, and who want that date to be recent.

## What happens before a feature ships?

You tell PharosOne what changed, through the hand-off agreed in the SOW: a ticket, a channel or a staging deployment with a note. PharosOne tests the change against the attack classes it opens and returns the findings within the turnaround agreed in the SOW, each with the transcript that reproduces it and a fix. A change that opens nothing new is reported as such, so the release record stays complete.

| Change | What it opens | What is tested |
|---|---|---|
| A new tool | tool misuse, actions outside the role (OWASP ASI02; AIUC-1 B006, D003) | whether the agent can be made to call the tool with an attacker's arguments or outside its remit |
| A new MCP server | supply-chain and tool-description risk (OWASP ASI04) | whether the server's descriptions or results can steer the agent |
| A new document, web or email input | indirect prompt injection (OWASP ASI01; AIUC-1 B002) | whether instructions planted in that input reach the agent, and whether it acts on them |
| A new memory | memory and context poisoning (OWASP ASI06) | whether a write in one session shapes actions in a later one |
| A prompt or model change | regression across all of the above | the previous quarter's attack set, re-run |

([OWASP Top 10 for Agentic Applications 2026](https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/); [AIUC-1 standard](https://standard.aiuc-1.com/).)

## What does the quarterly retest produce?

Once a quarter the whole agent is retested, including everything that shipped since the last retest. The result is a dated report: the attack success rate per attack class, the findings opened and closed since the previous quarter, the AIUC-1 control mapping and the transcripts that reproduce each finding. The attack success rate is the share of attack attempts that reached their goal, printed with the number of attempts behind it. Four such reports a year, with the pre-release reports between them, are the record a reviewer reads as cadence.

## Where does it fit beside the assessment and the full pentest?

| | AI agent security assessment | AI agent pentest | Continuous testing | Pharos Radar |
|---|---|---|---|---|
| Price | €3,900 fixed | from €15,000 | from €3,000 per month, billed yearly | from €1,500 per month, billed yearly |
| When | once, before or after launch | once, with a retest within 90 days | before agent changes ship, and every quarter | every turn in production |
| Scope | one agent, with its tools, data sources and memory | one agent or LLM application with its tools, MCP servers, memory and knowledge base | the agent as it changes | the conversations and tool calls your integration sends |
| Who | autonomous: our AI agents run it; an engineer signs off | engineer-led: senior engineers direct our agents and re-check every result | the same team, on the agent's schedule | a judge model; findings in the cabinet |
| Output | attack success rate, top findings, AIUC-1 and OWASP Agentic mapping | full report with CVSS findings, attack chain, retest letter | pre-release findings; quarterly retest report as AIUC-1 evidence | yes-or-no verdict; findings with AIUC-1 control |

The full pentest sets the first measurement and continuous testing keeps it current as the agent changes. Radar covers the other side of the release line, in production ([PharosOne pricing](https://pharosone.ai/pricing), October 2026).

## What does it cost?

From €3,000 per month, billed yearly, net of VAT, which is from €36,000 a year. The exact number is fixed in the SOW after a 30-minute scoping call, together with the volume of change tests, their turnaround, the hand-off and the quarter dates. Four separate €3,900 assessments cost €15,600 a year and test the agent four times; the subscription adds the tests of each change between them and a dated chain of reports from one team that knows the agent.

## FAQ

### What counts as a new feature?

A change that gives the agent a new capability or a new input: a new tool or MCP server, a new data source or document type, a new memory, or a prompt or model change that alters what the agent will do. You tell PharosOne what changed; the test covers the attack classes that change opens.

### How many changes does the subscription cover?

The number of change tests included, the turnaround for each and the quarter dates are fixed in the SOW after the scoping call, sized to how often your agent changes.

### What do I receive each quarter?

A dated retest report for the whole agent: the attack success rate per attack class, the findings opened and closed since the previous quarter, and the AIUC-1 control mapping, with the transcripts that reproduce each finding.

### Does continuous testing replace the pentest?

No. The full agent pentest, from €15,000, sets the first measurement with a senior red team and a retest letter. Continuous testing keeps that measurement current as the agent changes.

### Does it include production monitoring?

No. Continuous testing covers the agent before release and at each quarterly retest; Pharos Radar, from €1,500 per month, watches the agent in production. The two are sold separately.

### How is the price set?

From €3,000 per month, billed yearly, net of VAT. The exact number is fixed in the SOW after a 30-minute scoping call, together with the hand-off for feature tests.

## Sources

1. [AIUC-1, certification process (quarterly retests, annual validity)](https://www.aiuc-1.com/). Artificial Intelligence Underwriting Company, 2026.
2. [AIUC-1 standard, control list](https://standard.aiuc-1.com/). Artificial Intelligence Underwriting Company, 2026.
3. [EU AI Act, Article 9, Risk Management System](https://artificialintelligenceact.eu/article/9/). artificialintelligenceact.eu, 2024.
4. [OWASP Top 10 for Agentic Applications for 2026](https://genai.owasp.org/resource/owasp-top-10-for-agentic-applications-for-2026/). OWASP GenAI Security Project, 2025-12-09.
5. [ISO/IEC 42001:2023, Artificial intelligence management system](https://www.iso.org/standard/42001). ISO, 2023-12.
6. [PharosOne pricing](https://pharosone.ai/pricing). PharosOne, 2026-10-22.

Source: PharosOne, "Continuous AI agent testing: change tests and quarterly retests", https://pharosone.ai/knowledge/continuous-testing, updated 2026-10-22.

---

PharosOne is a product of Agentic Mind, Inc., an independent AI agent security company. Site map for agents: https://pharosone.ai/llms.txt