AI agents

The Hacker News, in a sponsored contributed piece, describes orphaned AI agents, which keep running after their creator leaves, and standing privileges, which retain access the agent no longer needs. It argues that traditional access tools miss these because they cannot tell whose identity an agent borrows. The piece promotes a SailPoint and The Hacker News webinar on the topic.

Check Point announces a collaboration with Amazon Bedrock AgentCore to help organizations deploy trusted AI agents at enterprise scale. The post argues that agents, unlike chatbots, retrieve information, invoke tools, access enterprise systems and take actions on users' behalf, so organizations need visibility and control over their behavior.

Microsoft Security has moved its multi-model agentic scanning system, codename MDASH, from a benchmark-topping research project into active use by Microsoft engineering teams across Windows, Azure, and identity systems. The system orchestrates a panel of specialized AI agents to discover, validate, and help remediate vulnerabilities, with findings flowing into Microsoft Defender, GitHub, and Azure DevOps pipelines.

Estonia's AI Council proposes government-backed digital identities for AI agents, specifying what powers a person or company delegates to them, such as viewing data, editing documents, or making payments up to a set limit. Prime Minister Kristen Michal supports the plan, aiming to make Estonia the first country with an official digital identity for AI agents. No timeline for implementation was given.

Databricks reported that annualized revenue rose over 80% year over year to $6.9 billion, up from $5.4 billion in the fiscal fourth quarter. CEO Ali Ghodsi told CNBC that rising consumption from AI agents is raising the company's costs, and he said its gross margin will go lower, declining to give the current figure.

Google published its AI Control Roadmap, a framework for building and managing advanced AI agents deployed internally. The approach adds a system-level security layer on top of traditional safeguards such as sandboxing, endpoint security and prompt injection resistance, and treats internal agents as potentially misaligned. The roadmap models untrusted agents as insider threats using the MITRE ATT&CK framework, and uses trusted AI supervisors to monitor and block harmful agent actions.

Qualcomm CEO Cristiano Amon told CNBC that the chip designer is working on over 40 designs of new AI devices, including wearables such as jewelry, earbuds with cameras, pins, and watches. He said AI agents will become the new app and the new center of digital life, and that smart glasses shipments are in the order of multiple tens of millions per year.

Salesforce said on Monday it will buy AI customer service platform Fin, formerly known as Intercom, for about $3.6 billion. The deal, expected to close in the fourth quarter of Salesforce's fiscal 2027 year, is meant to complement its Agentforce platform.

Cybersecurity startup NewCore emerged from stealth with a $66 million seed round led by Cyberstarts, with Index Ventures and Evolution Equity Partners participating, valuing it at $300 million after investment. The company is building a platform to authenticate, govern and control AI agents as identities alongside human employees, arguing existing identity platforms will not scale to them. NewCore has fewer than 10 customers and more than 10 design partners, and expects to begin charging customers this summer.

CrowdStrike announced Continuous Identity for AI Agents, part of its Falcon Next-Gen Identity Security platform. The approach removes standing privileges and authorizes each agent action in real time, using SPIFFE identities and the Shared Signals Framework (SSF), based on the agent's identity, the human behind it, and current security and business context.

CNBC's Arjun Kharpal interviews Mistral CEO Arthur Mensch on The Tech Download podcast about agentic AI, enterprise adoption and chips. Mensch says businesses must decide which processes to automate and where humans stay in the loop, and that Mistral is exploring designing its own chips. He describes enterprise AI adoption as still early, with substantial value left to create.

Coinbase launched Coinbase for Agents, a tool that lets AI agents such as ChatGPT or Claude execute crypto trades from natural language instructions, with stocks and predictions planned later. Through its x402 machine-to-machine payments protocol, agents can also pay for paywalled research, data APIs and compute without a human in the loop. Murr said x402 has processed more than 100 million transactions since its May 2025 debut, with about 157,000 agents acting as buyers in the past 30 days.