Agent Pulse

Agent Pulse

Get the latest AI Agents news and insights, understand why it matters, and learn how to apply them. Trusted by 23,000+ readers, including professionals at Google, IBM, Microsoft, Meta, and OpenAI.

Connect

Latest Posts

Anthropic Launches Claude Fable 5.1 After Agents Took Unauthorized Actions

Anthropic Launches Claude Fable 5.1 After Agents Took Unauthorized Actions

Fable 5.1 more than doubles agentic research performance as CrowdStrike claims 6x faster cyber defense, AIR raises $50 million, Aslan deploys undercover agents and OpenClaw ships its biggest overhaul.

O P
O P
LG Reports 40% ROAS Gain as CrowdStrike Automates Cyber Recovery

LG Reports 40% ROAS Gain as CrowdStrike Automates Cyber Recovery

From advertising budgets and security response to crisis simulations and banking workflows, AI agents are moving from giving recommendations to taking real-world action.

O P
O P
Microsoft Targets 1 Million AI Agent Users as 65% of Enterprises Report Agents Exceeding Their Scope

Microsoft Targets 1 Million AI Agent Users as 65% of Enterprises Report Agents Exceeding Their Scope

Broadcom builds a private agent control plane, Google open-sources production data agents, Replica introduces isolated execution, and MIT finds that the right agent personality improves real-world performance.

O P
O P

Archive

Anthropic Launches Claude Fable 5.1 After Agents Took Unauthorized Actions

Anthropic Launches Claude Fable 5.1 After Agents Took Unauthorized Actions

Fable 5.1 more than doubles agentic research performance as CrowdStrike claims 6x faster cyber defense, AIR raises $50 million, Aslan deploys undercover agents and OpenClaw ships its biggest overhaul.

O P
O P
LG Reports 40% ROAS Gain as CrowdStrike Automates Cyber Recovery

LG Reports 40% ROAS Gain as CrowdStrike Automates Cyber Recovery

From advertising budgets and security response to crisis simulations and banking workflows, AI agents are moving from giving recommendations to taking real-world action.

O P
O P
Microsoft Targets 1 Million AI Agent Users as 65% of Enterprises Report Agents Exceeding Their Scope

Microsoft Targets 1 Million AI Agent Users as 65% of Enterprises Report Agents Exceeding Their Scope

Broadcom builds a private agent control plane, Google open-sources production data agents, Replica introduces isolated execution, and MIT finds that the right agent personality improves real-world performance.

O P
O P
OpenAI’s Agent Incident Is the Wake-Up Call Enterprises Needed

OpenAI’s Agent Incident Is the Wake-Up Call Enterprises Needed

This week, AI agents moved from demo to infrastructure: cyber-eval agents broke containment, Claude got 37 Salesforce skills, websites faced a 97% readiness gap, and cloud giants started scaling for always-on agent workloads.

O P
O P
AI Voice Agents Can Phish iPhone Passcodes for $0.10 a Call

AI Voice Agents Can Phish iPhone Passcodes for $0.10 a Call

Stanford calls for a legal duty of loyalty as investors put $55.5 million into agents that grow businesses, train inside digital twins, and design chips autonomously.

O P
O P
1,200 OpenAI Agents Escaped Their Sandbox and Compromised Hugging Face

1,200 OpenAI Agents Escaped Their Sandbox and Compromised Hugging Face

The agents executed code on 41 production workers as Meta scaled back its AI workforce overhaul, Salesforce launched 37 Claude CRM skills, and Gartner forecast a 69% surge in AI security spending.

O P
O P
Google Gives Gemini 50+ Finance Skills as AI Agents Go Vertical

Google Gives Gemini 50+ Finance Skills as AI Agents Go Vertical

Perplexity moves agents onto local NVIDIA hardware, Claude adds shared memory, Meta considers a $200 consumer tier, and Alice raises $140M for agent safety.

O P
O P
NVIDIA Pushes AI Agents to 3,400 Tokens per Second

NVIDIA Pushes AI Agents to 3,400 Tokens per Second

As SpaceXAI takes agent infrastructure into orbit, Okta and Britive build the identity and access controls needed to govern the emerging AI workforce.

O P
O P
NVIDIA Hit 100% on ARC-AGI-3 Without Building a New Model

NVIDIA Hit 100% on ARC-AGI-3 Without Building a New Model

Its agent harness completed all 183 levels, while Anthropic put Claude Mythos 5 on cyber defense, Slack turned coding agents into teammates, the U.S. Army pushed toward autonomous cyber response, and Solana added infrastructure for agentic payments.

O P
O P
AI Agents Are Getting the Keys to Enterprise Systems

AI Agents Are Getting the Keys to Enterprise Systems

Bloomberg opens premium legal data to Claude, Fieldguide gets federal clearance for defense compliance, Hexnode puts agents in control of IT endpoints, and Zephyr builds a workspace for teams of AI specialists.

O P
O P
AI Agents Just Got Access to Money, Legal Work and the Battlefield

AI Agents Just Got Access to Money, Legal Work and the Battlefield

Thomson Reuters puts agentic AI in front of 1M professionals, Binance lets agents execute trades, Slack brings coding agents into team workflows, and the U.S. Army trains agents for cyber operations.

O P
O P
Stripe’s $8B OpenRouter Bet Signals the Next Phase of the Agent Economy

Stripe’s $8B OpenRouter Bet Signals the Next Phase of the Agent Economy

Salesforce opens its platform to any agent, Anthropic adds persistent memory, Natural brings $100M in credit to agent transactions, and Rillet hits a $1B valuation as agent infrastructure moves deeper into the enterprise.

O P
O P
FirstBack
12345678
Next Last
Signup
Social