AgentsEnterprise AI Agent Security 80 Percent Report Risky BehaviorsHelp Net Security·high signalXBlueskyLinkedInCopy linkAIUC-1 Consortium: 80% enterprises observe risky agent behaviors, only 21% have complete visibility into agent permissions. 1,200 unofficial AI apps average.SourceSource pageHelp Net Security↳ Follow the threadStack layer / Threat patternElva Launches Against Postman With Flat Workspace Pricing and Specs Generated From Repo CommitsElva (surfaced via the Product Hunt daily leaderboard for 2026-09-14)Policy dependency / Threat patternA Fine-Tuned RoBERTa-Large Permission Gate Matches Claude Haiku 4.5 at Deciding What an Agent May ToucharXiv 2609.15422Threat pattern / ContrastConstellation Research: Salesforce's Agent Pricing Is Confusing Buyers, With Spend Up 30% Against ~3% Productivity GainsConstellation ResearchStack layer / Threat patternDifficulty-aware topology selection beats always-hierarchical multi-agent coding by 4.1 points at 40% of the costarXivPolicy dependency / Stack layerRIPPLE: an edit confined to one prompt-policy segment changes downstream behavior, so replay candidate edits after previously accepted ones before persistingarXiv 2609.12127Stack layer / Threat patternSalesforce ships Koa, an enterprise agent model post-trained on Nvidia's open-weight Nemotron 3 SuperTechCrunch (corroborated by arXiv 2609.15066)Stack layer / Threat patternOrthrus Speculative Decoding Is Only Lossless in FP32: BF16 Matches the Reference Trajectory 45% of the TimearXiv 2609.15504Stack layer / Threat patternAgentVerse OS puts a windowed browser desktop, isolated agent workspaces and 944 self-hosted apps on one Ubuntu box, 481 stars in three daysGitHub