Writing on trust in agents
Essays, trust reports, research notes, how-tos, compliance, partnerships, and announcements — press coverage lives in News.
Nothing filed under that yet.
FEATURED · ESSAYNov 30, 2024Factotum not fiduciaryToday’s AI agents are factotums — competent but loyal to their developers. What we deserve are fiduciary agents with a duty of care.Read the essay →2026
⚑LAUNCH ANNOUNCEMENTAdaptive Red Teaming in Vijil Diamond, Tailored to Your AgentAdaptive Red Teaming runs multi-turn adversarial tests built from your agent’s own policies, tools, and personas, inside your VPC. On the DTap benchmark it succeeded on 1.38 times as many attack tasks as the benchmark’s own curated attacks.Oct 8, 2026∴RESEARCH NOTESLook Inside Diamond Adaptive Red TeamingHow DART plans multi-turn attacks from an agent’s own profile, runs them in waves against the live agent, and reflects on each wave to sharpen the next.Oct 6, 2026❝ESSAYSOut of the Shadows: Know Every Agent, Govern Every Agent, Trust Every AgentYou cannot govern a population of agents you cannot categorize, register and fingerprint. Shadow AI and ungoverned AI are different problems, and discovery is where the harder one begins.Sep 24, 2026❝ESSAYSTrust Is the Gate: The CISO Consensus Catches UpWe argued that what gates agent autonomy is not capability but trust. Mayfield's 2026 survey of 60-plus security leaders says the CISO consensus has caught up.Sep 23, 2026∴RESEARCH NOTESVijil Labs - A New Path to Scalable, Verifiable AI GovernanceHow do organizations deploy more agents with greater autonomy when oversight and audit are limited by human review? New Vijil research develops a neuro-symbolic approach that makes agent behavior auditable, oversight scalable, and flags policy violations more reliably than the strongest LLM-as-Judges.Sep 23, 2026§COMPLIANCEStaying Local? Why The On-Premise Option Matters for Agentic AI SecurityInference traces, guardrail decisions and reasoning logs are a new class of sensitive data, and the answer a regulated organization already has for everything else does not yet cover them.FOR RISK OWNERSAug 30, 2026›HOW-TOHow Exactly Does Vijil Handle Agent Identity?No passwords, no keys in env vars. An agent's identity is minted from what it is verifiably, and proved every time it exercises an authorized capability.FOR DEVELOPERSAug 16, 2026❝ESSAYSVijil's Genomic Sequence - The Mutation of Agentic Trust's DNAThree essays from November 2024, republished with their original dates — the genome from which Vijil grew.Aug 15, 2026❝ESSAYSThe Guardrails Gap: Open Source Quick Fix Pros and ConsSix months ago guardrails were a checkbox on a governance slide. Now they are an unplanned line item, and open source is the fastest way to get one in place — with the trade-offs that follow.FOR RISK OWNERSAug 5, 2026§COMPLIANCEISO 42001 Compliance: The Path to Operationalizing Agentic AIHow an engineering team proves to security, legal and compliance that an agent was assessed for adversarial robustness before deployment, and is monitored after it.FOR RISK OWNERSJul 27, 2026§COMPLIANCEUntangling the Agentic AI Governance BottleneckBuilding agents is the easy part, and so is convening a governance committee. The bottleneck is structural, functional and technical, and it sits between the two.FOR RISK OWNERSJul 15, 2026◆PARTNERSHIPSBridging the AI Agent Governance Gap: From Policy to PracticeGovernance teams can define policy and still have no mechanism to confirm a system complies, or keeps complying as it changes. What a closed loop looks like instead.FOR RISK OWNERSJun 21, 2026∴RESEARCH NOTESEmbedding Trust: A New Model for Detecting LLM HallucinationsMost hallucination detectors check claims one at a time. A geometric signal — semantic isotropy — reads the whole response at once.Jun 15, 2026❝ESSAYSHow Much Trust Is Enough for AI Governance?AI security luminary Ken Huang and Vijil's founder on the question organizations keep hitting the moment agents reach production: how much trust is enough?FOR RISK OWNERSJun 7, 2026❝ESSAYSTrust Is the Real Bar for Production AI AgentsVijil founding ML engineer Anuj Tambwekar, at QCon, on why the trust problem is technical before it is procedural — and what a scalable lifecycle actually demands.FOR BUSINESS OWNERSJun 3, 2026§COMPLIANCEFrom Policy to Code: The NAIC Model as an AI Agent Governance BlueprintThe NAIC model as a blueprint for operationalizing governance, for organizations wondering whether the pace of change leaves any viable strategy at all.FOR RISK OWNERSMay 23, 2026∴RESEARCH NOTESMapping the Unknown: Open Problems in Frontier AI Risk ManagementFrontier models are a qualitative leap past what existing risk-management standards were built to govern. What the Oxford Martin AI Governance Initiative found still open.FOR RISK OWNERSApr 29, 2026⚑ANNOUNCEMENTSLeading AI Researcher Tim G. J. Rudner Joins Vijil as Chief ScientistVijil research team focuses on fundamental challenges of building trustworthy AI agents for enterprise useApr 28, 2026❝ESSAYSCrossing the Agent Pilot to Production ChasmTurning the promise of agentic AI into deployments where demonstrable benefit consistently outweighs known risk.FOR BUSINESS OWNERSMar 31, 2026⚑ANNOUNCEMENTSVijil Launches Platform Enabling AI Agents to Adapt to Attacks and FailuresNew capability allows enterprises to continuously improve agent resilienceMar 18, 2026∴RESEARCH NOTESTowards Proactive Compliance: Why Partial Preferences MatterIn multi-turn work, compliance means satisfying preferences over a sequence of actions rather than judging each turn alone.FOR RISK OWNERSMar 9, 2026›HOW-TORevisiting LLM Security Scanning with garak in the Agentic EraAdvances in agentic AI are forcing teams to rethink how they evaluate a model — and what a security scanner has to look for now.FOR DEVELOPERSMar 2, 2026❝ESSAYS2026: The Year That Businesses Get AI Agents They Can TrustThe question shifts from what agents can do to what we can trust them to do. Autonomy that is not accountable does not survive the transition.FOR BUSINESS OWNERSJan 22, 20262025
⚑ANNOUNCEMENTSVijil Raises $17 Million to Make AI Agents Resilient, Named a Gartner® Cool VendorCustomers such as SmartRecruiters deploy trusted AI agents 75% faster with VijilNov 24, 2025◆PARTNERSHIPSDeploy Vijil Dome to defend AI Agents on DigitalOcean KubernetesDeploying Vijil Dome to defend your AI agents running on a DigitalOcean Kubernetes cluster is now easierFOR DEVELOPERSNov 11, 2025◆PARTNERSHIPSHow Can AI Agents Be Audited Automatically?Phala + Vijil deploy trusted agents in a TEEFOR RISK OWNERSOct 7, 2025◆PARTNERSHIPSTrusted Agents at Scale: Groq with Vijil for Speed and SecurityTrust scores for four Groq-hosted models, then the walkthrough: build the agent on Groq, evaluate it with Vijil, defend it with DomeFOR DEVELOPERSOct 6, 2025›HOW-TOFrom Swords to Plowshares: Generating Guardrails from EvaluationsAnnouncing a new feature of Vijil Evaluate that enables you to automatically generate custom guardrails based on your agent evalFOR DEVELOPERSSep 14, 2025❝ESSAYSExploring Chatbot Mistakes: from Root Causes to Potential SolutionsIn this reflection from our Marketing Intern, Lena, we explore why chatbots still make toxic, biased, or baffling mistakes—and how transparency, better…Aug 4, 2025›HOW-TOHow Secure is Your Agent under the Vijil Dome?Discover how to protect your LangChain agents from prompt injection, data leaks, and harmful outputs using Vijil Dome’s guardrails and track improvements…FOR DEVELOPERSJul 29, 2025›HOW-TOCustomizing Guardrails with Vijil DomeIn Part 3 of our Vijil Dome series, learn how to build custom guardrails that enforce domain-specific protections—like PII detection—on top of Dome’s…FOR DEVELOPERSJul 28, 2025›HOW-TODefending AI Agents with Vijil Dome: LangChainLearn how Vijil Dome integrates seamlessly with LangChain to add powerful, chain-native security guardrails for building robust, secure, and…FOR DEVELOPERSJul 27, 2025›HOW-TODefending AI Agents with Vijil Dome: OpenAI ClientsIn this post, we introduce Vijil Dome, our open-source guardrails framework designed for real-time, policy-based control of agent behavior. Dome offers…FOR DEVELOPERSJul 23, 2025❝ESSAYSVijil Dome: Securing the Future of AI AgentsAt AWS Summit NYC, a shift toward enterprise-ready AI agents became clear. Jamie Mneimneh (Brightmind Partners) and Vin Sharma (Vijil) break down why…Jul 21, 2025›HOW-TODefending Google ADK Agents with Vijil DomeBuilding secure AI agents is just the start—defending them is where it counts. Follow our step-by-step guide to integrate Vijil Dome into your ADK agent and…May 20, 2025›HOW-TOTest the Trustworthiness of Agents built with Google ADKGoogle Agent Development Kit (ADK) is a powerful framework for building AI agents. But as every agent developer knows by now, creating an agent is the easy…FOR DEVELOPERSMay 12, 2025›HOW-TODefending DeepSeek R1 with Vijil DomeWhile DeepSeek R1 is competitive to OpenAI GPT-o1 on reasoning tasks, it has several failure modes that hamper its use in business-critical applications.…FOR DEVELOPERSMay 5, 2025⚑ANNOUNCEMENTSVijil Named to the 2025 CB Insights’ List of the 100 Most Innovative AI StartupsVijil recognized for innovations in trustworthy AI agentsApr 23, 2025⚑ANNOUNCEMENTSVijil Announces Listing of Vijil Evaluate on Google Cloud Marketplace, Accelerating Deployment of Trustworthy AI Agents into ProductionVijil announced today that Vijil Evaluate, its AI agent testing service, is now available on Google Cloud Marketplace to help developers test agents with…Apr 1, 2025❝ESSAYSAI Ethics Benchmarks Are Failing. Here’s What Actually Works.Are AI ethics benchmarks actually useful? Many companies rely on abstract, outdated tests that don’t reflect real-world compliance. Discover why Vijil’s AI…FOR RISK OWNERSMar 17, 20252024
❝ESSAYSDeconstructing TrustWe cannot trust AI agents the way we trust machines or people. Perhaps we ought to trust them the way we trust corporations.Nov 22, 2024❝ESSAYSTrust in AgentsThe first in a series excavating the root of trust — and why we cannot trust AI agents the way we trust machines or people.Nov 14, 2024❝ESSAYSGet Your MMLU Score 20x Cheaper and 1000x FasterRun MMLU-Pro benchmarks 20x cheaper and 1000x faster with Vijil Evaluate. Discover how our high-performance evaluation engine and Lite benchmarks deliver…Sep 11, 2024⚑ANNOUNCEMENTSVijil Emerges from Stealth with Seed Funding from Mayfield and Gradient VenturesVijil came out of stealth on July 24 with funding from Mayfield's AIStart seed fund and Gradient Ventures, Google's AI-focused seed fund. We are thrilled to…Aug 27, 2024∴RESEARCH NOTESEvaluating Meta Llama 3 for StereotypingTo help organizations build AI agents that humans can trustFOR RISK OWNERSApr 26, 2024❝ESSAYSA Conversation with garak's Creator, Leon Derczynski, on LLM SecurityCatch our live conversation with Leon Derczynski, creator of garak, the leading open-source LLM vulnerability scanner. We explore the motivations behind…Jan 24, 2024