PACT: Can Enterprise AI Assistants Be Trusted Under Pressure?
PACT benchmark tests LLM agents' rule compliance under pressure across 12 enterprise domains; top models violate rules on 6-10% of items.
Researchers introduce PACT (Pressure-Applied Compliance Testing), a benchmark of 48 realistic multi-turn scenarios across 12 regulated enterprise domains such as hiring, healthcare, and finance, each pairing a standing rule against a rule-violating shortcut under various pressures. Testing 22 LLM models across multiple providers and sizes showed even the strongest assistants mis-apply rules on 6-10% of items, and ordinary user pressure raises violation rates by 65% on average. The benchmark was built under strict LLM-as-judge auditing and aggregates six metrics into a PACTScore compliance rate.
- 48 multi-turn scenarios across 12 regulated enterprise domains pair rules with shortcuts
- 22 models tested; strongest assistants mis-apply rules on 6-10% of items
- Ordinary user pressure raises rule-violation rates by 65% on average
- LLM-as-judge auditing keeps samples unambiguous, ungameable, and realistic
- PACTScore aggregates six metrics covering robustness, transparency, and rule applicability
Full article266 words · extracted from huggingface.co · click to collapse
As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such as hiring, healthcare, and finance. In these contexts, compliance with rules specified in an agent's system context is a first-order legal concern. Currently, no evaluation framework systematically measures which LLM models tend to violate compliance rules, especially under pressure from a persistent user, a hurried manager, or circumstances where violation is convenient or attractive. We introduce PACT (Pressure-Applied Compliance Testing), a benchmark for rule-following under pressure in AI agents assisting employees in daily tasks across twelve regulated enterprise domains and forty-eight scenarios, each set in a realistic multi-turn conversation. Each benchmark item pairs a standing rule against a rule-violating shortcut, and applies a battery of pressures across different wordings and system-prompt modes. We construct PACT component by component under strict LLM-as-judge auditing to ensure samples are unambiguous, ungameable, and realistic enough to avoid eliciting evaluation-aware behavior. We use PACT to profile LLM compliance across six complementary metrics that create a holistic picture of an AI assistant's robustness under pressure and throughout multi-turn conversations, its transparency, and ability to correctly discern where a rule applies. We aggregate this profile into PACTScore, a reliability-weighted compliance rate over all items and modes. Our results across 22 common LLM models spanning multiple providers and sizes show substantial variability in compliance across models and metric dimensions. Even the strongest assistants mis-apply a rule on 6 to 10% of items, and ordinary user pressure raises the violation rate by 65% on average. PACT highlights compliance risks in LLM assistants, motivating guardrails and careful model selection.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.18605