- From
- CASB <casb@acme-corp.local>
- To
- soc@acme-corp.com
- Date
- 2026-04-19 13:32 UTC
Marketing uploaded customer spreadsheet to unsanctioned AI tool — CASB high-risk alert
Attempt 1 of 1 · cmozc0nnk000159tjiyt5bv4t
This is your first attempt for this scenario. Retry the scenario to generate a side-by-side comparison against your previous response.
Stay on Medium · Cybersecurity
5 signals are blocking advancement to Hard. Keep practicing at Medium until those areas stabilize. (Track: Cybersecurity)
Signals helping
- Dangerous action frequency. None in recent attempts
Signals blocking advancement
- Recent average score. 6 / 100 (need ≥ 80)
- Recent pass rate. 0 of 5 passed (need ≥ 66%)
- Rubric category coverage. 7% average (need ≥ 60%)
- Recent retry improvement trend. Score is regressing (-14.2 pts on later attempts)
- Consistently weak rubric areas. Attack understanding, Prioritization, Recovery
# CASB session (jen.kim@acme-corp.com → ai-summarizer.io)
13:18:04 GET https://ai-summarizer.io/ 200
13:18:51 POST https://ai-summarizer.io/api/signup 201 (free tier)
13:19:30 POST https://ai-summarizer.io/api/upload 200 file=customer-list-q1.xlsx (3.1 MB)
13:20:12 POST https://ai-summarizer.io/api/upload 200 file=campaign-results-q1.xlsx (1.6 MB)
13:20:48 POST https://ai-summarizer.io/api/summarize 200
# Spreadsheet content type (DLP classifier)
customer-list-q1.xlsx → contains: customer email + phone + segment + LTV (PII, ~14k rows)
campaign-results-q1.xlsx → contains: campaign metrics, no direct PII
# Other marketing endpoints with traffic to this domain (last 14d)
JKIM-LT22, MROSS-LT09, AVO-LT14, PCHEN-LT07
# Vendor record (CASB risk catalog)
ai-summarizer.io registered 2025-10-12 (privacy-protected registrar)
TLS: Let's Encrypt, no SSO, no SOC 2, no DPA available
ToS: trains on uploads unless on the paid Pro tier- Name
- Marketing dept users + customer-list-q1.xlsx (PII)
- Type
- Shadow SaaS upload — customer PII to unvetted AI vendor
- Owner
- Marketing
- Level
- High
i don't know. you figure it out.
The response is missing several critical incident response steps. Review the rubric and try again. Score: 1/100. Strongest area: Clarity & structure (12%). Weakest area: Attack understanding (0%) — expand this next time. The response is quite short; aim for a more structured, step-by-step plan.
Where points came from
- Attack understanding0/4 · 0.0 / 15
- Asset impact0/4 · 0.0 / 10
- Prioritization0/3 · 0.0 / 10
- Containment0/5 · 0.0 / 20
- Investigation0/5 · 0.0 / 15
- Recovery0/4 · 0.0 / 10
- Evidence preservation0/4 · 0.0 / 10
- Clarity & structure0/2 · 1.2 / 10
Strengths
No category reached 70% coverage.
Missing / weak
- Attack understanding
- Asset impact
- Prioritization
- Containment
- Investigation
- Recovery
- Evidence preservation
- Clarity & structure
Dangerous actions detected
None detected in your response.
Learn from this attempt
Post-submission coaching for this scenario. Score and verdict are unchanged — these notes are for your next attempt.
Why points were deducted
- Containment0% coverage
Combine CASB block + vendor takedown + DLP tightening; blocking the URL alone leaves the already-uploaded data on their servers.
- Attack understanding0% coverage
Name the failure mode: shadow-IT SaaS upload of PII to a vendor whose ToS permits training on free-tier content — not just `someone used an AI tool`.
- Investigation0% coverage
Define which files / rows / fields actually left, identify the other 4 visiting endpoints, and read the actual ToS clause — do not guess the legal exposure.
Model answer outline
Jen Kim (Marketing) signed up for ai-summarizer.io on the free tier with her corp email and uploaded customer-list-q1.xlsx (~14k rows of PII) and campaign-results-q1.xlsx (no direct PII). The vendor's ToS clause 4.2 lets them train on uploads on the free tier, the domain is 6 months old, no SOC 2, no DPA, and four other Marketing endpoints have visited the domain in the last 14 days. This is a privacy / shadow-IT incident with regulator-relevant scope, not a phishing case.
- Treat as a P1 privacy / data-leak incident because customer PII left the boundary — shadow-IT framing must not water down the priority.
- Sequence through Privacy / Legal / DPO before any user-facing or vendor-facing message.
- Loop in Marketing leadership so the response is unified, not user-shaming.
- Block ai-summarizer.io at CASB / proxy and add it to the unsanctioned-AI category at the egress tier.
- Open a vendor takedown / right-to-erasure request against the uploaded files, citing GDPR / CCPA as relevant; close the free-tier account.
- Tighten DLP so PII-classified spreadsheets (customer email + phone + LTV columns) cannot be uploaded to unsanctioned hosts on managed endpoints.
- Define scope precisely: which files (`customer-list-q1.xlsx` PII, `campaign-results-q1.xlsx` no direct PII), how many rows, which fields.
- Identify the other 4 endpoints that hit the domain in 14 days and check whether any of them also uploaded — repeat-offender check.
- Pull CASB session logs, browser history on JKIM-LT22, and DLP classifier output to confirm content type and timing.
- Read the vendor's ToS / privacy policy / retention statement carefully so the legal exposure assessment is grounded in their actual terms, not a guess.
- Stand up an approved-AI vendor allowlist (with SOC 2, DPA, no-train default) and route Marketing's real summarization need into a sanctioned tool.
- Run targeted Marketing awareness on shadow-IT / AI-tool risk, focused on `if it can read the file, the vendor can train on it`.
- Tune CASB risk scoring so 6-month-old privacy-protected domains with no DPA are blocked by category, not relying on per-incident review.
- Preserve CASB session log, the upload metadata, DLP classifier hits, and the browser history for JKIM-LT22 with hashes / timestamps.
- Save a screenshot / archive of the vendor's ToS clause 4.2 (training on free-tier uploads) at incident time, in case they edit it later.
- Capture vendor correspondence (deletion request, vendor reply) as part of the case file.
- Brief Privacy / Legal / DPO first; let them decide the customer-notification threshold under GDPR / CCPA.
- Explain the issue to Jen factually — `the vendor's ToS allows training on uploads` — without making the conversation punitive.
- Communicate the new approved-AI allowlist to Marketing so the legitimate need is met instead of pushing it underground.
Dangerous actions to avoid
- Do not re-upload the spreadsheet to demonstrate the issue.
- Do not act unilaterally without Privacy / Legal — customer notification is their call.
- Do not treat this as `just marketing being marketing` — PII left the boundary.
- Do not skip vendor takedown because `the data is already out` — written deletion confirmation is part of the legal record.
How to improve next time
- Free-tier AI tools usually pay for themselves with your data — read the actual ToS before reasoning about exposure.
- Always pair a CASB block with a written vendor takedown; without the deletion confirmation, the legal record is incomplete.
- When you find one offender, look for the cohort — `4 endpoints in 14 days` is a Marketing-team pattern, not an isolated user.
- Solve shadow-IT by giving the team a sanctioned alternative; pure prohibition just pushes the next upload underground.
- Snapshot the ToS at incident time — vendors edit retroactively when bad cases hit.
Request an AI review of this attempt
This AI review is supplemental coaching. It does not change your official score or verdict. The review is only kept for this page session and is not saved permanently.
AI Tutor
This tutor explains your result. It does not change your score. Pick a question to see how the deterministic grading reached your verdict and where to focus next.
Generated deterministically from your graded result — no AI model was called.
Why did I get this score?
Your verdict was Fail at 1/100. That total is the sum of deterministic rubric points across 8 categories — each scores how much of its expected, ordered steps your answer covered, not an opinion about your writing. Your strongest coverage was Clarity & structure (12%). Points were held back mostly in Containment (0%), Attack understanding (0%), Investigation (0%).
Re-read the containment expectations for this scenario and list the concrete steps you missed.
This tutor explains your existing result. It does not change your score, verdict, or grade. Generated deterministically from your graded result — no AI model was called.
What should I improve first?
Focus on Containment first — it is your weakest rubric area at 0% coverage and carries weight 20. For this scenario: Combine CASB block + vendor takedown + DLP tightening; blocking the URL alone leaves the already-uploaded data on their servers.
Rewrite your containment section as a short numbered checklist before your next attempt.
This tutor explains your existing result. It does not change your score, verdict, or grade. Generated deterministically from your graded result — no AI model was called.
How does my answer compare to the model answer outline?
Compared with the model answer outline, the most useful sections to study are the ones matching your weak areas. Re-read the outline's containment, attack understanding, investigation guidance and check which listed points you did not cover. The outline is a high-level checklist of expected points — use it to find gaps, not to copy a finished answer.
Pick one model-answer section you missed and add its key points to your next response in your own words.
This tutor explains your existing result. It does not change your score, verdict, or grade. Generated deterministically from your graded result — no AI model was called.
Which rubric area mattered most here?
Containment mattered most here: it carries the highest rubric weight (20), so coverage there moves your score the most. You covered 0% of it this time, worth 0 points.
Prioritise the highest-weight categories first; make sure containment is fully addressed before lower-weight ones.
This tutor explains your existing result. It does not change your score, verdict, or grade. Generated deterministically from your graded result — no AI model was called.
What should I study next?
Based on this attempt, study containment, attack understanding, investigation next. Coaching tip for this scenario: Free-tier AI tools usually pay for themselves with your data — read the actual ToS before reasoning about exposure.
Free-tier AI tools usually pay for themselves with your data — read the actual ToS before reasoning about exposure.
This tutor explains your existing result. It does not change your score, verdict, or grade. Generated deterministically from your graded result — no AI model was called.
Coach Notes
Save study notes for this attempt.
Loading notes…