Filings, briefings, and event recaps from AFI's ongoing effort to educate policymakers on the opportunities and risks of agentic AI.
A plain-language guide for congressional staff and regulators, covering what an AI agent is, how these systems work, the core vocabulary, and the policy questions they raise. No technical background required.
The same guide as a printable, shareable document for briefing books and email. Covers the shift from answering to acting, the core vocabulary, the autonomy spectrum, and the policy questions that follow.
A plain-language inventory of who is building the standards for agentic AI and what they are, spanning the Linux Foundation's Agentic AI Foundation, formal bodies like W3C and the OpenID Foundation, NIST, and the emerging commerce and payments layer.
AFI's policy brief making the pro-competition case for agentic AI, covering cloud marketplaces, enterprise-grade business intelligence, and data integration platforms as leveling forces for small businesses and startups.
Nine pages covering threat taxonomy, identity frameworks, risk tiering, and deployment environment controls for agentic AI systems. Submitted in response to the CAISI Request for Information on AI Agent Security.
AFI and Cravath, Swaine & Moore host a fireside chat in New York on October 20 with David Shaw, Principal Deputy Director of the FTC Bureau of Competition, and Levi Swank, Deputy Director for Litigation in the FTC Bureau of Consumer Protection. The conversation covers how agentic AI is reshaping markets, business practices, and the FTC's priorities, followed by a reception.
AFI convened congressional staff, federal and international officials, and industry leaders in Washington for an invitation-only dinner on agentic AI in healthcare, moderated by AFI Founding Executive Director Ryan Dattilo. The discussion ranged from where agents can deliver meaningful value in care delivery and administration to the identity, oversight, and patient protections needed to build trust.
AFI's inaugural symposium convened industry leaders and lawmakers on the policy questions around deploying AI agents, featuring Reps. French Hill, Zach Nunn, and Ted Lieu, with speakers from Anthropic, AWS, Intuit, Hippocratic AI, GoDaddy, Imbue, Palo Alto Networks, EdgeRunner AI, and the FDA.
AFI Founding Executive Director Ryan Dattilo briefed the G7 Cyber Expert Group, convened by the Bank of Italy in Palermo, on agentic AI and the next phase of cybersecurity for the financial sector. He presented alongside Hart Rossman, VP of Global Services Security at AWS and an AFI member, pairing a policy perspective on the considerations raised by wider agentic deployment with a practitioner's view of how frontier models are reshaping vulnerability detection and defense.
Senior officials from the VP's office, OCC, Treasury, NIST, UK FCA, and European Parliament join AFI members to discuss agentic AI in financial services.
Brought together founding members, other industry players, Members of Congress, representatives from the Department of Energy, and the British Embassy.
AFI formally launches as a cross-industry coalition to educate policymakers on agentic AI, with founding members across AI development, cloud infrastructure, financial services, defense, and enterprise.
In an op-ed, AFI Founding Executive Director Ryan Dattilo draws on his role planning Mark Zuckerberg's first congressional hearing to examine how public comfort has shifted from sharing data to letting AI agents act on our behalf, and what that shift means for privacy, authorization, and accountability.
Rep. Lance Gooden argues Congress should lead on agentic AI through oversight rather than premature regulation, bringing together innovators, legal experts, and industry to address liability and accountability questions before courts and state laws define the field by default.
The Information reports on Sen. Mark Warner's forthcoming discussion draft to create rules for AI agents, covering duty-of-loyalty obligations, platform interoperability, and FTC and NIST oversight roles. The piece names AFI as a D.C. coalition convening industry and lawmakers on agent policy, including at its recent symposium.
Punchbowl News covers AFI's inaugural symposium, where the coalition convened industry leaders and lawmakers on the policy questions around deploying AI agents, highlighting the tech industry's push to get Congress up to speed on agentic AI.
At SAP Connect in Las Vegas, SAP introduced Joule Work, a framework for assigning tasks to agents across SAP and third-party data through natural language, along with new Joule Agents for product design, supply planning, manufacturing, logistics, and asset maintenance that can connect to third-party agents and systems.
AWS introduced a managed runtime for building stateful agents with OpenAI models on Amazon Bedrock, with all inference running inside AWS and Bedrock AgentCore as the default compute environment. AWS says authorization policy enforcement, agent discovery, and observability and evaluation tools are coming next.
SailPoint combined its Human Fabric and Agentic Fabric products into a single identity security system and made Agentic Fabric generally available. Agentic Fabric discovers AI agents, credentials, and MCP servers that sit outside traditional security tools, and adds controls such as a centralized kill switch for rogue agents.
GoDaddy is helping develop Agent Name Service, an open standard for agent identity, naming, and verification built on DNS and public key infrastructure, alongside Infoblox's DNS-based approach to agent discovery. The two companies argue agent identity should be open and interoperable rather than tied to proprietary registries.
At AWS Summit New York, AWS announced general availability of the Bedrock AgentCore managed agent harness, letting teams stand up a production agent in two API calls, alongside new capabilities connecting agents to organizational, web, and paid knowledge, production failure and trajectory insights, and governance controls that scale as agents grow more capable.
Plaid partnered with Fin, an AI agent for customer service, to bring the Plaid Link experience directly into the Fin AI Agent, letting fintechs and financial institutions help users securely connect or reconnect bank accounts inside a chat without leaving the conversation, part of Plaid's shift from digital finance toward intelligent finance.
Anthropic unveils 10 pre-built AI agents for banking, insurance, asset management, and fintech, alongside Claude Opus 4.7, its most capable model for financial work, in a push to become the operating layer for Wall Street.
Anthropic launches Claude for Small Business, with prebuilt agentic workflows and connectors embedded directly into QuickBooks, PayPal, HubSpot, Canva, DocuSign, and Microsoft 365, putting AI agents within reach of any small business owner.
AWS introduces Amazon Bedrock AgentCore Payments, letting autonomous AI agents make real-time purchases using stablecoins, in partnership with Coinbase and Stripe, as a foundational building block for the agentic economy.
Plaid's State of Intelligent Finance Report (Spring 2026) examines how AI is reshaping what consumers and businesses expect from financial technology.
AWS introduces AgentCore, a new set of capabilities in Amazon Bedrock designed to accelerate enterprise AI agent development and deployment at scale.
Anthropic introduces Claude Managed Agents to streamline and accelerate AI agent development for enterprise teams building production-grade agentic applications.
EdgeRunner AI launches WarClaw, an agentic decision system designed for real-time autonomous operations at the tactical edge for defense applications.
Intuit and Anthropic announce a multi-year partnership to deliver trusted financial intelligence and custom AI agents to consumers and businesses through Intuit's platform.
Unstructured published a detailed enterprise blueprint explaining why most AI agent pilots fail due to poor data architecture, with 70–85% of AI project failures stemming directly from data issues.
Lilt published findings on how global content leaders are using agentic AI for translation at scale.
EdgeRunner released military-specific AI agents for logistics, combat medicine, and combat arms at no cost to Department of War users.
Plaid detailed two internal AI agents: the AI Annotator for data labeling and Fix My Connection for repairing bank integrations, enabling over 2 million successful user-permissioned logins and reducing fix time by 90%.
AWS unveiled AgentCore Memory, AgentCore Policy, and AgentCore Evaluations at re:Invent 2025, expanding its suite of tools for building, governing, and assessing enterprise AI agents.
Anthropic unveiled its enterprise agents program, its most aggressive push to integrate agentic AI into everyday workplaces, including prebuilt plugin templates for financial analysis, investment banking, equity research, and engineering.
Members receive monthly AI policy briefings, advance notice of events, and access to all AFI filings and publications.