AIxSafe Banking
CPS 234 Telemetry Dashboard
Real-Time Payments • CPS 234 Focused

Real time AI oversight for regulated banking workflows

Surface sensitive prompt exposure, policy effectiveness, inspection coverage and review-ready evidence across ChatGPT, Claude, Copilot and other approved AI tools. Built to support information asset protection, incident visibility and control assurance.
Total AI prompts
6,542
Past 30 days
Sensitive prompts
451
Detected by DLP inspection
High risk users
14
Exceeded daily threshold
Control evidence coverage
97%
Reviewable records retained
AI Sensitive Data Exposure Rate
6.9%
Within pilot tolerance
Sensitive prompts detected ÷ total prompts. This is the lead board metric for CPS 234 oversight.
Inspection coverage
96.8%
AI traffic inspected vs detected
Decision trace rebuild
< 4 min
Average investigation time
Critical exposure events
12
High confidence multi-field matches
Unique AI users
41
Across monitored devices

Real-Time Payments & AI Telemetry

The tiles below are the banking telemetry board mapped into CISO reporting views. Overview shows the full operating picture, while the other buttons filter to the most relevant cards for each lens.

AI Influenced Decisions+0
4,200

Payment workflows where AI-assisted decision support was recorded across the pilot period.

Policy Enforcement Checkpoints+0
11,800

Policy evaluation events applied across transaction lifecycle stages during the engagement.

Anomaly Escalation Events+0
38

Operational incidents where telemetry signals triggered manual review or elevated intervention.

Decision Trace Reconstruction
< 4 min

Average time to rebuild a complete AI decision pathway during post-incident investigation.

Vendor Routing Transitions+0
640

Workflows recorded interacting with more than one automation or AI provider in a single transaction journey.

Exception Handling Load+0
312

Workflow exceptions requiring operational team involvement outside automated processing paths.

Telemetry Evidence Coverage
97%

Proportion of AI-influenced workflows with retained reviewable telemetry records at pilot close.

Governance Review Requests+0
14

Requests from risk and compliance functions for decision-level telemetry during the engagement period.

Board level CPS 234 metrics

Focused on information asset exposure, incident visibility and control effectiveness. These are the numbers risk and compliance teams should be able to explain at any point in time.

AI Sensitive Data Exposure Rate

6.9%
+0.8% vs prior period

High Risk AI User Rate

16.7%
3 more users above threshold

AI Inspection Coverage

96.8%
+2.1% after policy tuning

Evidence Retention Coverage

97.0%
Complete for most workflows

Sensitive prompt timeline

Distribution of low, medium and high confidence matches across the last 12 reporting intervals.

Peak interval 13:00 43 matches flagged across payment operations prompts
High confidence share 18% Above baseline due to customer data and credential patterns
Escalation posture Contained 5 intervals crossed the manual review threshold
Total sensitive matches per hour Medium confidence contributes to rising load High confidence drives review escalation

Sensitive data categories

Breakdown of detected exposures across categories relevant to information asset protection.

Financial information
42%
Customer identifiers
27%
Source code
18%
Credentials and secrets
13%
Internal documents
9%

AI application risk map

Sensitive prompts by approved AI application.

ApplicationPromptsSensitive
ChatGPT2,860214
Copilot1,74096
Claude1,12074
Perplexity51012
Gemini3127

High risk users

Users with repeated sensitive prompt events above the defined threshold.

UserEventsRisk
ops.payments18High
fraud.review15High
service.team.311Medium
engineering.api8Medium
risk.analyst.26Medium

Critical exposure events

High confidence prompts containing multiple sensitive fields or higher consequence data types.

Event typeCountStatus
Multi-field financial data5Open
Credential plus code exposure3Reviewing
Customer identifier bundle2Reviewing
Internal document fragment2Closed

Control effectiveness and evidence

Operational measures that map inspection, monitoring and retained evidence to the control expectations around ongoing testing and information security capability.

Prompts scanned by DLP
On target
6,335
Traffic inspected and evaluated against active detection logic.
Prompt evidence retained
Strong
97.0%
Reviewable encrypted records retained for oversight and incident investigation.
Bypassed AI traffic
Needs attention
3.2%
AI requests detected outside full inspection path due to exemptions or endpoint variance.
Average trace reconstruction
Fast
< 4 min
Time to rebuild an AI decision path using retained telemetry and decrypted evidence.

Sample dashboard for local review. Metrics are illustrative and structured for CPS 234 conversations.