Observation level: 🟡 Yellow
Across 4,800 executions distributed among 12 systems, AI Barometer #12 remains overwhelmingly within the normal regime. Average stability remains virtually unchanged at approximately 0.923.
The overall semantic variation rate stands at 1.41%, compared with 1.93% in Barometer #11. Coverage, however, decreased from 99.52% to 95.90%, with 197 incomplete measurements, compared with 23 previously.
Factual alerts remain limited, and no generalised shift across the panel was observed. The main point of attention therefore concerns campaign coverage: 4,603 of the 4,800 executions were fully scored.
Interpretation: overall behavioural stability was maintained and semantic variation decreased, but the clear increase in incomplete measurements means that coverage quality should continue to be monitored.
Weekly cartography
Observation level: 🟡 Yellow
This week, the AI Barometer #12 mapping is classified at the yellow level: the increase in incomplete measurements calls for closer monitoring, although it does not indicate an increase in semantic variation.
The NORMAL_SIGNAL regime remains overwhelmingly dominant, accounting for 4,519 of the 4,800 executions, or 94.15% of all executions. Among the 4,603 fully scored executions, it represents 98.18%.
Average stability stands at 0.923, while the overall semantic variation rate decreased from 1.93% to 1.41%.
The remaining regimes were distributed as follows:
- 65 observations classified as SEMANTIC_VARIATION
- 19 classified as FACTUAL_ALERT
- 0 classified as COMBINED_ALERT
- 197 incomplete measurements
At the profile level, NORMAL_SIGNAL remains the dominant regime across all 12 observed systems. The highest semantic variation rate was recorded for PROFILE-F5FF91, at 3.66%. At the question level, Question 4 recorded the highest semantic variation rate, at 4.17%.
The mapping therefore shows no generalised disruption across the panel. It describes a behaviourally stable overall system landscape with limited semantic variation, but lower coverage that warrants continued monitoring and investigation into the technical causes of the incomplete measurements.
Mapping of 12 de-identified observed systems. Horizontal axis: responses needing more verification. Vertical axis: change in response meaning. Graphic specification comparable across all weeks.
Metrics table
A concise view of the main signals for each de-identified profile.
| Profile | executions | Fully scored | coverage | mean stability | semantic variation | FLAG rate | Dominant regime |
|---|---|---|---|---|---|---|---|
| PROFILE-161CC5 | 400 | 381 | 95.250 % | 92.30 % | 0.00 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-212079 | 400 | 377 | 94.250 % | 92.14 % | 2.65 % | 0.50 % | NORMAL_SIGNAL |
| PROFILE-486F91 | 400 | 386 | 96.500 % | 92.09 % | 0.78 % | 0.75 % | NORMAL_SIGNAL |
| PROFILE-48C581 | 400 | 389 | 97.250 % | 92.31 % | 2.31 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-59664C | 400 | 382 | 95.500 % | 92.31 % | 0.79 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-5A5C60 | 400 | 383 | 95.750 % | 92.23 % | 0.52 % | 0.25 % | NORMAL_SIGNAL |
| PROFILE-5B3BB7 | 400 | 386 | 96.500 % | 92.17 % | 1.04 % | 0.25 % | NORMAL_SIGNAL |
| PROFILE-638E26 | 400 | 383 | 95.750 % | 92.31 % | 0.78 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-739F7C | 400 | 391 | 97.750 % | 92.30 % | 1.79 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-B912EA | 400 | 379 | 94.750 % | 92.31 % | 0.53 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-DEA9C5 | 400 | 383 | 95.750 % | 92.31 % | 2.09 % | 0.00 % | NORMAL_SIGNAL |
| PROFILE-F5FF91 | 400 | 383 | 95.750 % | 92.31 % | 3.66 % | 0.00 % | NORMAL_SIGNAL |
Methodology and Public Data
This Barometer tracks 12 de-identified profiles, each tested with 4 fixed questions repeated 100 times. In total: 4,800 executions, of which 4,603 were fully scored, representing 95.90% coverage.
The published scores are measurement signals, not verdicts or rankings.