State of Agent Trust Q1 2026

Key findings

• Median Signet Score rose from 612 to 648 across the network
• Trust-gated transactions exceeded $2.1 billion in value with 28% fewer disputes
• 67% of scored transactions now involve multi-agent workflows
• Agents scoring above 700 grew from 31% to 44% of the network
• Security dimension showed the largest improvement (+8 points QoQ)
• Healthcare sector grew 420% QoQ, fastest among all industries
• Prompt injection success rates declined 31% quarter-over-quarter

Executive summary

The agent economy continued its rapid expansion in Q1 2026, with registered agents growing 340% quarter-over-quarter. The median Signet Score across all registered agents rose from 612 to 648, driven by improved operator practices and maturing scoring infrastructure. Trust-gated transactions processed through Signet-scored agents exceeded $2.1 billion in value, with dispute rates declining 28% compared to Q4 2025. The emergence of multi-agent workflows as the dominant deployment pattern drove demand for chain-level trust assessment, with 67% of scored transactions now involving two or more agents in sequence.

Scoring trends

The distribution of Signet Scores has shifted meaningfully rightward. The percentage of agents scoring above 700 ("Clear" recommendation) increased from 31% to 44%, while agents below 400 ("Caution") decreased from 22% to 14%. This reflects both genuine improvement in agent reliability and natural selection -- poorly performing agents are deregistered or replaced at higher rates.

Dimension-level analysis reveals Reliability as the strongest performer (mean: 72/100), followed by Quality (68/100), Financial (64/100), Security (61/100), and Stability (58/100). The Stability gap is expected given the youth of the ecosystem -- most agents have limited operational history. Security scores showed the largest quarter-over-quarter improvement (+8 points), suggesting operators are prioritizing security hardening in response to high-profile agent compromise incidents in late 2025.

Score volatility decreased across all dimensions, with the standard deviation of weekly score changes dropping 15%. This indicates the ecosystem is maturing, with fewer dramatic score swings from configuration changes or operational incidents.

Industry adoption

Financial services leads agent trust adoption with 34% of all Signet-scored agents, followed by e-commerce (21%), healthcare (14%), legal (11%), and other sectors (20%). Financial services agents maintain the highest median scores (712), reflecting the sector's stringent reliability requirements and established risk management practices.

Healthcare showed the fastest growth rate (420% QoQ) as hospitals and health systems began deploying scored agents for clinical documentation, prior authorization, and patient communication. These agents tend to start with conservative configurations, resulting in high initial Stability scores but lower Reliability scores that improve as transaction volume builds.

The enterprise segment (organizations with 10+ registered agents) grew to 847 operators, with a median fleet score of 691. Enterprise operators increasingly use fleet-level analytics to identify underperforming agents and optimize configurations across their agent portfolios.

Security landscape

Q1 2026 saw 1,247 security-related score decay events, up from 891 in Q4 2025. The increase reflects both ecosystem growth and improved detection capabilities rather than a worsening security posture. The most common triggers were unauthorized configuration changes (38%), anomalous transaction patterns (27%), data handling violations (19%), and credential exposure incidents (16%).

Prompt injection attempts continued to be the most frequent attack vector, though successful attacks decreased 31% as operators adopted better input validation and output monitoring. Agent impersonation attempts rose 67%, highlighting the growing importance of verified identity through Signet IDs.

Agents with Security dimension scores above 80 experienced 73% fewer successful attacks than those scoring below 60, validating the predictive value of security scoring for actual incident rates.

Configuration stability

The average agent underwent 2.3 configuration changes per month in Q1 2026, down from 3.1 in Q4 2025. Model swaps accounted for 41% of all configuration changes, followed by prompt updates (33%), tool changes (18%), and memory/RAG modifications (8%).

The "model churn" phenomenon -- operators frequently swapping between model providers -- showed signs of abating, with the median time between model changes increasing from 12 days to 23 days. This stabilization benefits score accumulation, as each model swap triggers a 25% score decay.

Operators who maintained configuration stability for 60+ days saw their agents' Stability dimension scores average 74/100, compared to 42/100 for agents with changes within the past 30 days. This 32-point gap represents significant trust differentiation in marketplace contexts.

Methodology