This Month's AI Agents News

Nearly 9 in 10 companies will have AI talking to customers by year's end. Can it tell when you've had enough?

AI agent failures make the news regularly. One of the best-known examples is probably [DPD’s AI assistant](https://www.itv.com/news/2024-01-19/dpd-disables-ai-chatbot-after-customer-service-bot-appears-to-go-rogue), which in 2024 failed to track a missing parcel, then failed to connect the customer with a human representative and to give him a phone number for customer service. In the [Sinch survey](https://sinch.com/ai-production-paradox/chapter/ai-production-challenges/), 22% of organizations running live AI agents had rolled one back due to hallucinations or brand risk. ### “I don’t know” might be the smartest answer Across all organizations running AI agents in customer communications, [74% have done it at least once](https://sinch.com/ai-production-paradox/chapter/ai-production-challenges/). Which means the first signal telling a company its AI agent has failed is somebody complaining to customer service or on social media.

Amazon blocks Meta's AI Agent Muse AI on its app, unhappy with Facebook-parent capturing and storing cust

Tech News News: Amazon has reportedly blocked Meta's new autonomous personal assistant, Muse, from navigating and purchasing products on Amazon.com on...

XMR Price Revival Puts Focus on MemeToro’s Transparent AI Launchpad Model and Fair-Launch Smart Contract Code

Monero is trading near $580 after breaking above a descending trendline that had limited its rallies since January. This change has placed $585.37 in focus,...

Raindrop announces series A and $50M in total funding Led by CRV to protect the World from AI agent failures

# Raindrop announces series A and $50M in total funding Led by CRV to protect the World from AI agent failures September 22, 2026  [](https://www.island-oil.com/?utm_source=cyprusshippingnews&utm_medium=banner&utm_campaign=islandoil-csn) _The Series A will accelerate Raindrop’s anomaly detection research, scale the product to more enterprises, and advance Simulations, a new product that tests agent changes against production traffic before they ship_ Raindrop, the agent reliability company, raised a Series A led by CRV, bringing total funding to $50 million. Existing investors Lightspeed Venture Partners, Y Combinator participated along with lead researchers from OpenAI, Anthropic, and Thinking Machines. Raindrop is trusted by some of the largest and fastest-growing companies in the world, including Vercel, Framer, Clay, and Fortune 100 Enterprises to protect their users from agent failures. Simulations replay real production traffic and existing test cases against a proposed change to an agent harness, then apply Raindrop’s anomaly detection to the results. In other words, AI engineers can now see “what their change will change.” Agents are becoming more capable and complex. The research group METR found that the length of tasks agents can complete on their own doubles roughly every seven months, and a single run can now last days and involve thousands of tool calls. “Agents now run for hours, call thousands of tools, and handle real money, real health data, and real customers,” said Zubin Koticha, CEO of Raindrop. Raindrop catches those failures in production, and with Simulations, before a change ever ships. This funding lets us bring the testing process frontier labs use on their own models to every

How AI Agents Can Trigger Runaway Costs for Enterprises

# How AI Agents Can Trigger Runaway Costs for Enterprises ## Want more Dark Reading stories in your Google search results? - [Essential News & Insights from Black Hat USA 2025](https://dr-resources.darkreading.com/c/pubRD.mpl?qf=w_darl25&ch=mod) ### You May Also Like [Application Security\ \ **Supply Chain Attack Secretly Installs OpenClaw for Cline Users** \ \ by Rob Wright\ \ Feb 19, 2026](https://www.darkreading.com/application-security/supply-chain-attack-openclaw-cline-users#recipe=similar-items&source_content_id=1d1c29eff49426577dcf3e84cf3b98ff) [Application Security\ \ **Chinese Hackers Hijack Notepad++ Updates for 6 Months** \ \ by Jai Vijayan\ \ Feb 02, 2026](https://www.darkreading.com/application-security/chinese-hackers-hijack-notepad-updates-6-months#recipe=similar-items&source_content_id=1e51d280261e34e0cc1db58fc4422dce) [Application Security\ \ **Trump Administration Rescinds Biden-Era Software Guidance** \ \ by Alexander Culafi\ \ Jan 29, 2026](https://www.darkreading.com/application-security/trump-administration-rescinds-biden-era-sbom-guidance#recipe=similar-items&source_content_id=5f19c8eeeb93473c11fd921316903dea) [Application Security![Black blocks and one red block showing a bug or glitch in a computer progr

How to Evaluate AI Agents From Tool Calls to Task Completion | NVIDIA Technical Blog

# How to Evaluate AI Agents From Tool Calls to Task Completion ## Analyzing [Nemotron 3.5 Lightning](https://developer.nvidia.com/blog/nvidia-nemotron-3-5-lightning-delivers-fast-accurate-specialized-task-execution-for-long-running-agents/) through this lens [Scroll to Analyzing Nemotron 3.5 Lightning through this lens section](https://developer.nvidia.com/blog/how-to-evaluate-ai-agents-from-tool-calls-to-task-completion/#analyzing_nemotron_35_lightning_through_this_lens) On **PinchBench**, Nemotron 3.5 Lightning hits 86% accuracy while finishing 10,000 tasks 30% faster than Qwen3.6 35B at comparable accuracy — a model that finishes efficiently beats one scoring higher on isolated accuracy while burning more steps and [tokens](https://blogs.nvidia.com/blog/ai-tokens-explained/). _On PinchBench, Nemotron 3.5 Lightning reaches 86% accuracy while completing tasks up to 30% faster than Qwen3.6 35B at comparable accuracy_

AI agents raise new safety questions

One area getting attention is AI agents. Ron Johnson and Congressman Tony Wied for their position on AI guardrails and have not heard back yet.

The advertising industry is building standards for a world of AI agents

For years, advertising technology has been built around systems designed for people to operate. Buyers search inventory, compare options, negotiate terms,...

Amazon Blocks Meta's Muse AI Agent From Shopping Platform

Meta's AI shopping agent Muse has been abruptly blocked from accessing Amazon's platform, marking the first major restriction on AI agents in e-commerce.

Amazon cuts off Meta's new AI shopping agent

Sherwood Protocol Puts AI Agent Funds On Robinhood Chain

# Robinhood Let AI Agents Trade. This Startup Wants Them Running Funds By [Boaz Sobrado](https://www.forbes.com/sites/boazsobrado/), Contributor. Forbes contributors publish independent expert analyses and insights. Boaz Sobrado is a fintech leader and entrepreneur. Follow Author Sep 21, 2026, 02:05pm EDT 0 [Add Forbes on Google](https://www.google.com/preferences/source?q=forbes.com) Listen to the article Robinhood Let AI Agents Trade. This Startup Wants Them Running Funds 0:00 / 6:00 1x 2x 1.75x 1.5x 1.25x 1x 0.75x Verbatim Verbatim Robinhood Let AI Agents Trade. This Startup Wants Them Running Funds This voice experience is generated by AI. [Learn more](https://www.forbes.com/sites/forbesdigitalgroup/article/forbes-ai-audio-learn-more/). This voice experience is generated by AI. [Learn more](https://www.forbes.com/sites/forbesdigitalgroup/article/forbes-ai-audio-learn-more/).  Close-up of a person's hand holding a smartphone and using the Opus 4 model within the Claude app from AI company Anthropic, Lafayette, California, May 22, 2025. (Photo by Smith Collection/Gado/Getty Images) Gado via Getty Images Robinhood opened its brokerage to AI agents in May. The design is cautious: a separate account, only the money you put in it, a notification for every trade and an off switch. Sherwood Labs, a startup building on Robinhood’s own blockchain, thinks the cautious version is the wrong shape. It wants agents to coordinate and manage capital in shared vaults and run them like funds, with token-staking “guardians” replaying every trade before it executes. Its WOOD token has traded on Robinhood Chain since July. The protocol itself still runs on a simulation of it. “We’re building Sherwood to be the capital coordination layer for agentic finance,” Carlos Beltran, co-founder of Sherwood Labs, said in an interview. The company is buildin

Seattle startups DexCare and Mila Health combine to expand AI patient scheduling

Seattle health-tech startup [DexCare](http://dexcarehealth.com/) today [announced](https://www.businesswire.com/news/home/20260921009868/en/DexCare-Acquires-Mila-Health-Extending-AI-Agents-From-Scheduling-to-Care-Coordination) it’s acquiring Mila Health, a fellow Seattle-area company that has built AI agents to call and text patients seeking medical appointments. The startup said it takes only hours to embed its technology into existing systems and launch AI agents. “With DexCare, providers and payers can configure an AI agent to complete the tasks they can’t staff for.” The startup appears not to have raised any venture capital and has 25 employees.

Are AI Agents Actually Saving Money? How CTOs Should Measure Agentic AI ROI

As enterprises deploy agentic workflows, token prices may be falling, but overall AI spending is still increasing.

Arm Surges 13% as Meta's Muse Agent Reignites CPU Demand Bet; Intel Jumps 12%, AMD Climbs 9%

# Arm Surges 13% as Meta’s Muse Agent Reignites CPU Demand Bet; Intel Jumps 12%, AMD Climbs 9% **Meta Platforms** ( [NASDAQ:META](https://247wallst.com/companies/META/?tpid=1664811&tv=link&tc=in_content) | [META Price Prediction](https://247wallst.com/companies/meta/price-prediction?tpid=1664811&tv=link&tc=in_content)) and its Muse artificial intelligence agent are driving the move, signaling that [inference workloads](https://247wallst.com/investing/2026/05/12/amd-ceo-lisa-su-says-cpus-will-match-gpus-1-to-1-in-ais-next-phase/?tpid=1664811&tv=link&tc=in_content) are ramping faster than expected and that autonomous AI agents could materially raise CPU footprint per unit of compute. ## Meta’s Muse Agent Reignites the CPU Bet Autonomous AI agents sharply increase the need for inference, the process of running a trained model against live data, and inference work requires CPUs alongside accelerators for orchestration and data movement.

KeewanoDB stores events for agent analysis

# KeewanoDB stores events for agent analysis Israeli startup [Keewano](https://www.keewano.com/) has a database storing data for machine reasoning at scale by AI agents, and has revealed $12 million in total seed funding. Its CEO and co-founder Mark Kardashov said: “We were trying to use AI agents to answer business questions like ‘Which of our users are on the same path as the ones who churned last month?' AI agents need a fundamentally different data architecture, and we’ve built KeewanoDB as the new line.”

Corridor raises $25M seed to build a health benefits brokerage for SMBs

# Corridor raises $25M seed to build a health benefits brokerage for SMBs Later, he teamed up with former Scale AI colleague Eric Qian to work on Capernaum AI, a startup building a clinical agent targeting musculoskeletal care and chronic pain.

Cognizant: How the CISO Role Evolves in the Age of AI Agents

# Cognizant: How the CISO Role Evolves in the Age of AI Agents It is no secret that indeterminate systems like [autonomous agents](https://cybermagazine.com/news/uk-firms-cant-see-their-own-ai-agents-cohesity-warns) have changed the cybersecurity landscape as we knew it.

Best Memecoin Launchpads Utilize AI Agents and Smart Contracts to Enforce Fair Launches on BNB Chain

- [🎯The AI Agents Creates A Checkable Launch Proposal](https://bitcoinfoundation.org/news/ai/memecoin-launchpads-utilize-ai-agents-and-smart-contracts-to-enforce-fair-launches-on-bnb-chain/#the-ai-agents-creates-a-checkable-launch-proposal)

How V7 gives AI agents institutional memory

# How V7 gives AI agents institutional memory —Alberto Rizzoli, Co-Founder and CEO at V7 ## Choosing OpenAI to keep complex workflows on track V7 also tested GPT‑6 Astra on its most difficult queries. ## Keep reading [View all](https://openai.com/news/) [Higgsfield AI ships new video features in a day with GPT-6 Astra\ \ StartupSep 21, 2026](https://openai.com/index/higgsfield-from-prompt-to-production-with-astra/)

CrowdStrike warns AI agents are creating a cyber battleground for GCC organisations

## AI agents introduce new security risks “AI has moved from being an information provider via chat to full autonomy and agency,” said Daccache. “Now we have AI agents that are much faster, much more autonomous, work 24/7 and never sleep. ## The race against machine-speed attacks “George Kurtz, our CEO, recently said that the concept of breakout time is dying – attacks today are becoming instant,” Daccache said, adding that autonomous AI agents can execute commands, move laterally across environments and achieve attacker objectives in seconds. “With AI agents gaining full autonomy, they can execute scripts and move laterally within seconds,” he said.

Kastle Raises $24 Million to Help Lenders Adopt Agentic AI

# Kastle Raises $24 Million to Help Lenders Adopt Agentic AI By[PYMNTS](https://www.pymnts.com/author/pymnts/) | September 21, 2026 [Share on Facebook](https://www.facebook.com/sharer.php?u=https://www.pymnts.com/news/artificial-intelligence/2026/kastle-raises-24-million-to-help-lenders-adopt-agentic-ai/&t=Kastle+Raises+%2424+Million+to+Help+Lenders+Adopt+Agentic+AI)[Tweet This Post](https://twitter.com/share?text=Kastle+Raises+%2424+Million+to+Help+Lenders+Adopt+Agentic+AI%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20-&url=https://www.pymnts.com/news/artificial-intelligence/2026/kastle-raises-24-million-to-help-lenders-adopt-agentic-ai/)[Share on LinkedIn](https://www.linkedin.com/shareArticle?mini=true&url=https://www.pymnts.com/news/artificial-intelligence/2026/kastle-raises-24-million-to-help-lenders-adopt-agentic-ai/&title=Kastle+Raises+%2424+Million+to+Help+Lenders+Adopt+Agentic+AI&summary=&source=)[Share via Email](mailto:?subject=Kastle%20Raises%20$24%20Million%20to%20Help%20Lenders%20Adopt%20Agentic%20AI&BODY=I%20found%20this%20article%20interesting%20and%20want%20to%20share%20it%20with%20you.%20Check%20it%20out:%20https://www.pymnts.com/news/artificial-intelligence/2026/kastle-raises-24-million-to-help-lenders-adopt-agentic-ai/) | Listen to ArticleListen  [Kastle](https://www.kastle.ai/), an agentic AI platform for consumer lenders, has raised $24 million in new funding. ## Get the Full Story Δ Alternative: WPA The company’s Series A round, [announced](https://www.prnewswire.com/news-releases/kastle-raises-24m-series-a-led-by-insight-partners-to-build-the-ai-workforce-for-banking-operations-302881290.html) last week and led by [Insight Partners](https://www.insightpartners.com/), will help Kastle expand its engineering, product and go-to-market teams, deepen its platforms’ capabilities and accelerate

UN panel calls for stronger safeguards as AI agents advance

# UN panel calls for stronger safeguards as AI agents advance [Facebook](https://www.facebook.com/sharer.php?u=https://news.un.org/en/story/2026/09/1168380) [Twitter](https://twitter.com/intent/tweet?text=UN%20panel%20calls%20for%20stronger%20safeguards%20as%20AI%20agents%20advance%20&url=https://news.un.org/en/story/2026/09/1168380) [Print](https://news.un.org/en/story/2026/09/1168380) [Email](mailto:?subject=UN%20panel%20calls%20for%20stronger%20safeguards%20as%20AI%20agents%20advance%20&body=https://news.un.org/en/story/2026/09/1168380) [The UN-backed Independent International Scientific Panel on AI](https://www.un.org/independent-international-scientific-panel-ai/en)’s warning followed the hack of the online platform HuggingFace between May and July by “AI agents” during a test initiated by OpenAI, the company behind ChatGPT.

Binghamton projects awarded AI Seed Grant funding to advance teaching, research, and public engagement

Eight Binghamton University research projects have been awarded AI Seed Grant funding to advance teaching, research, and public engagement.

AI Agents Take the Controls in Supply Chains

# AI Agents Take the Controls in Supply Chains ## Get the Full Story An autonomous agent detects rising security risks around the Bab el-Mandeb Strait and, rather than waiting for human approval, evaluates satellite imagery, maritime intelligence, weather conditions, freight rates and inventory positions across multiple continents.

AI agent startup Health Force raises €4.2 million

[Skip to content](https://htworld.co.uk/news/ai/ai-agent-startup-health-force-raises-e4-2-million-htai26/#content) # AI agent startup Health Force raises €4.2 million By [News Editor](https://htworld.co.uk/author/newseditor/)Published On: September 21, 2026Last Updated: September 21, 2026  ### **Health Force has raised €4.2m in Seed funding to expand its hospital [AI agents](https://htworld.co.uk/leadership/are-ai-agents-the-new-answer-to-drug-development-delays-lead26/) into more workflows and grow its engineering and delivery teams.** The Barcelona-based health tech company develops AI agents designed to automate administrative and operational tasks in hospitals. Founded in Vienna in 2022 by Fadi Haddad and Dr Juan Suárez, Health Force says its technology works alongside hospital operations teams, autonomously managing processes under human supervision. [](https://aspectpublishing.shorthandstories.com/health-tech-world-awards-2026-entry/) Haddad, co-founder and chief executive of Health Force, said: “Hospitals don’t need another dashboard, they need the work done. “We offer hospitals pre-trained AI agents that log into existing systems and complete real work. “All agents run on the same platform, which means hospitals work with one partner to solve many problems. “It also means overcoming Europe’s market fragmentation: the same platform automating hospital-insurance workflows in Italy automates SICA reporting in Portugal, and aims to automate KHVVG reporting in Germany next.” The agents use their own computers and logins to access systems already used by hospitals and carry out tasks within those platf

Why practical beats polished if you’re going to solve startup problems

# Why practical beats polished if you’re going to solve startup problems ## Get the best of Startup Daily straight to your inbox ### **6 lessons from creating [Spark: Startup Growth Lab](https://luma.com/sparkstartupgrowthlab)** We built the workshops with a clear problem at the centre, so the audience could bring their actual one — their AI agent, their sales problem, their CFO questions — and have it answered there and then by the experts in the room.

Ant International Launches Full-Stack AI-Native Solutions for Global Business Operations

News, Technology · September 21, 2026. Peng Yang, CEO of Ant International, introduced the company's payment foundation model at Ant.

Bumrungrad turns to AI agents to clear patient email backlog

Bumrungrad International Hospital is using Salesforce's Agentforce to summarise and route patient correspondence, with automated appointment booking and...

Amazon blocks Meta’s Muse AI assistant in new standoff over agentic shopping

Amazon blocked Meta’s new Muse personal AI agent from shopping on Amazon.com — citing security, privacy, and transparency concerns — after attempting unsuccessfully to get the Facebook parent company to exclude the e-commerce site from the experience. As of Sunday night, people using Muse to shop on Amazon see the popup, “Continued access by an unauthorized AI agent violates Amazon’s Conditions of Use, to which our customers have agreed.” Amazon has spent the past year trying to keep outside agents off its site, [suing Perplexity over its Comet browser](https://www.retaildive.com/news/amazon-sues-perplexity-ai-shopping-agents/804871/) and moving to block shopping agents from [Google](https://www.tipranks.com/news/amazon-amzn-blocks-googles-ai-shopping-agents-from-its-e-commerce-platform) and [OpenAI](https://www.techradar.com/ai-platforms-assistants/chatgpt/amazon-blocks-chatgpts-new-shopping-agent-what-the-fallout-could-mean-for-you). **How Muse works:** Meta [launched Muse on Sept. 8](https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/) as a U.S.-only personal agent that carries out multi-step tasks rather than answering one-off questions, connecting to services including email, calendar, payments, dining and shopping.

Stealien AI agent tops APEX 2026 as Korea-led team outperforms human rivals - CHOSUNBIZ

Artificial intelligence (AI) security corporations Stealien said on the 21st that its autonomous AI security agent took first place at the international...