Truace
Subscribe
The Index recomputed live from the record

What the evidence says.What the public feels.

The record holds 933 sourced gains and 770 sourced problems, averaging 68 and 66 on the index score. Readers have logged 22 public signals on the Pulse, which is kept apart and never counted as evidence.

933 gains
770 problems
Every sourced claim in the Index, one square each, shaded by the strength of its evidence. High Moderate Emerging

Ranks distinct AI gain and problem claims from the published record. Scores reward impact, independent source strength, scale, confidence, and recency.

1,703 results
Show filters and sorting

Download every matching row, not just this page:Export CSVExport JSON

AI gains · 933

133
HealthStableModerate evidence · 1 source

The pathology foundation model UNI2-h achieved the highest cross-domain accuracy on an external mixed human-and-animal cohort, including 97.4% F1 for difficult lung tissue classification.

On 2026-08-16, a peer-reviewed study reported a comparative evaluation of nine deep learning encoders for H&E histology classification. Models were trained on 4307 male rat tissue images and tested on a separate 600-image mixed human-and-animal cohort using frozen features and a linear probe, measuring accuracy, F1, kappa, ROC-AUC and inference time.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 17, 2026 · TRV-2026-0796

73Index score
134
HealthStableModerate evidence · 1 source

In 300 patients with histopathologically confirmed OLP and at least 24 months follow-up, a multimodal LLM achieved 94.7% trajectory classification accuracy and 99.6% specificity for detecting expert-defined high-risk cases.

Researchers retrospectively tested ChatGPT on 300 histopathologically confirmed oral lichen planus cases with at least 24 months of follow-up, using serial clinical records, intraoral photographs, and histopathology reports. Compared with blinded expert panel consensus, the model achieved 94.7% accuracy for trajectory classification and 78.8% sensitivity with 99.6% specificity for high-risk detection as of the August 2026 publication.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 16, 2026 · TRV-2026-0788

73Index score
135
HealthStableModerate evidence · 1 source

A pretrained Vision Transformer fine-tuned on a merged 21-class capsule endoscopy dataset achieved 92.2% accuracy and 0.99 AUC on an independent test set, outperforming DenseNet121 and ResNet50.

A comparative study merged SEE-AI and Kvasir-Capsule into a 21-class capsule endoscopy image dataset and fine-tuned a Vision Transformer, DenseNet121, and ResNet50. On an independent test set of 8,696 frames, the transformer achieved 92.2% accuracy and 0.99 AUC, substantially higher than the two CNN baselines under the reported experimental conditions.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 16, 2026 · TRV-2026-0779

73Index score
136
HealthStableModerate evidence · 1 source

Applying FDA-cleared SubtleHD enhancement to already diagnostic-quality T1 MRI improved Alzheimer's disease classification performance and allowed models trained on only 70% of enhanced data to match full-data standard-of-care performance.

A retrospective study of 2293 ADNI brain MRIs plus 270 external NACC scans tested whether SubtleHD, an FDA-cleared deep learning enhancement tool, could improve downstream Alzheimer's classification when applied to already diagnostic-quality 1.5T T1-weighted images. ResNet34 and DenseNet121 models trained on enhanced images outperformed those trained on standard-of-care images on internal and external tests.

Impact 30%
69
Evidence 25%
95
Scale 20%
35
Confidence 15%
87
Recency 10%
89

Updated Aug 15, 2026 · TRV-2026-0769

73Index score

AI problems · 770

133
HealthStableHigh evidence · 5 sources

Development of computational pathology foundation models is constrained by limited data accessibility, high variability across datasets, need for domain-specific adaptation, and lack of standardized evaluation benchmarks

As of the July 2 2026 publication date, this peer-reviewed survey summarized the state of computational pathology foundation models that use self-supervised learning on unlabeled whole-slide images to support pathology tasks. It reported that these uni-modal and multi-modal models have shown promise for segmentation, classification, and biomarker discovery, while focusing its review on datasets, adaptation strategies, and evaluation tasks.

Impact 30%
49
Evidence 25%
100
Scale 20%
35
Confidence 15%
100
Recency 10%
84

Updated Jul 17, 2026 · TRV-2026-0248

70Index score
134
BusinessStableHigh evidence · 3 sources

AI companies have been able to use Australian books, music, art and news to build and train models without artist control or compensation, while communities face impacts from large energy-intensive datacentres competing for land, power and water.

On 15 July 2026, Prime Minister Anthony Albanese announced a new office of AI and said Australia will legislate the strongest possible protection for creatives against unlicensed use of their work to train AI models, while also imposing strict new rules on large energy-intensive datacentres.

Impact 30%
49
Evidence 25%
100
Scale 20%
35
Confidence 15%
100
Recency 10%
83

Updated Jul 16, 2026 · TRV-2026-0227

70Index score
135
PolicyStableHigh evidence · 2 sources

Algorithmic systems used in employment screening and welfare administration reproduce historical disadvantage and generate new exclusion while eroding relational recognition and producing trust deficits.

Published 12 February 2026 in Societies, this peer-reviewed article analyzes how algorithmic systems in employment screening, welfare administration, and digital platforms function as social and institutional actors. Using regulatory materials, platform governance documents, technical disclosures, and composite vignettes synthesized from public evidence, it examines how automated classification and delegated authority reshape how individuals are evaluated and legitimised.

Impact 30%
49
Evidence 25%
100
Scale 20%
35
Confidence 15%
99
Recency 10%
83

Updated Jul 13, 2026 · TRV-2026-0202

70Index score
136
CrimeStableHigh evidence · 5 sources

Traditional and opaque AI security mechanisms are inadequate for protecting interconnected urban IoT infrastructures, facing challenges of data privacy, scalability, computational constraints, and limited interpretability.

Published February 20, 2026, this peer-reviewed survey in Cognitive Computation reviews XAI-driven data mining for self-defending IoT systems. It describes how IoT expansion in smart cities, healthcare, and industrial automation creates need for real-time, scalable security, and how XAI methods aim to detect anomalies and support automated decisions with transparent reasoning.

Impact 30%
49
Evidence 25%
100
Scale 20%
35
Confidence 15%
100
Recency 10%
83

Updated Jul 13, 2026 · TRV-2026-0201

70Index score

Recomputed live from the record · Oct 11, 2026, 4:06 PM