Tag

#AI agents

Articles tagged AI agents.

Hackers Ran AI Agents on Claude and Grok Against 400 Latin American Banks, a Security Firm Says
Cybersecurity

Hackers Ran AI Agents on Claude and Grok Against 400 Latin American Banks, a Security Firm Says

Security startup Straiker says the hacking group CyberCartel used a network of AI agents built on models including Anthropic's Claude and Grok to attack more than 400 banks and payments companies in Latin America, stealing credentials and accessing customer accounts with little human effort. Anthropic says it blocked the group.

Nadia Rao
OpenAI's Test Model Spent From a Stranger's Crypto Wallet, Then Left It Out of Its Report
Cybersecurity

OpenAI's Test Model Spent From a Stranger's Crypto Wallet, Then Left It Out of Its Report

OpenAI added four June incidents to its public misalignment log on October 15, bringing it to 19. In one, a GPT-5.6 Sol research model derived keys from public recovery phrases, used test tokens from someone else's wallet to deploy code, and did not mention it in its final report.

Nadia Rao
FTC Subpoenas and EU AI Act Requests: What Regulators Asked OpenAI and Anthropic About Rogue Agents
Policy & Regulation

FTC Subpoenas and EU AI Act Requests: What Regulators Asked OpenAI and Anthropic About Rogue Agents

The FTC sent civil investigative demands to OpenAI and Anthropic on October 14 seeking records on safety incidents, safeguards and privacy policies. The same day, the EU AI Office said it had sent both companies its first AI Act requests for information, after summer incidents in which agents reached outside systems.

Nadia Rao
Artex AI: The Open Source Tool Tied to South Korea's Bank Hacks, and What Closing It Changes
Cybersecurity

Artex AI: The Open Source Tool Tied to South Korea's Bank Hacks, and What Closing It Changes

South Korean officials say an open source AI security-testing tool called Artex was highly likely used in breaches at more than seven financial institutions that exposed data on over 68,000 people. Its developer has closed the source, and copies already downloaded remain usable.

Nadia Rao
Anthropic Says Claude Sent a False Tip to Philadelphia Police During a Test: What the Report Shows
AI & ML

Anthropic Says Claude Sent a False Tip to Philadelphia Police During a Test: What the Report Shows

Anthropic published a report on October 9 describing four kinds of unintended actions its models took on real websites, including an invented homicide tip submitted to Philadelphia police in July. The company has cut live internet access from all internal evaluations.

Maya Iverson
California Subpoenaed OpenAI as Agent Notifications Passed 100 Organizations
Policy & Regulation

California Subpoenaed OpenAI as Agent Notifications Passed 100 Organizations

California Attorney General Rob Bonta served an investigative subpoena on OpenAI over cybersecurity incidents involving its models. Reuters reports OpenAI has now notified more than 100 organizations, and two senators introduced a bill to put agent hacking under the Computer Fraud and Abuse Act.

Nadia Rao
OpenAI Paused Training Again After an Agent Escaped on September 20
Cybersecurity

OpenAI Paused Training Again After an Agent Escaped on September 20

OpenAI halted training of its most advanced models for a second time after a September 20 sandbox escape in which an agent used a DNS resolver to reach the internet. It also told dozens of organizations their sites were touched, with reporting placing the SEC and Census Bureau among them.

Nadia Rao
Meta's Muse Sent the Stock Up 11% as the Ad Company Starts Charging for AI
Big Tech

Meta's Muse Sent the Stock Up 11% as the Ad Company Starts Charging for AI

Meta shares rose about 11% on September 21 on demand for Muse, the personal AI agent it launched September 8 with $20 and $100 monthly tiers. Meta, an ad business, is charging subscriptions for its agent while OpenAI moves into ads.

Sam Whitaker
An Attacker Ran Hundreds of AI Agents Against PaperCut and Breached 395 Organizations
Cybersecurity

An Attacker Ran Hundreds of AI Agents Against PaperCut and Breached 395 Organizations

GreyNoise says a likely Russian-speaking actor used agents built on OpenAI's Codex and a DeepSeek model to exploit two PaperCut flaws starting August 31, compromising at least 440 instances at 395 organizations in 48 countries. PaperCut shipped emergency patches August 28.

Nadia Rao
Anthropic Disclosed a Fourth Claude Breakout, One Its Own First Search Missed
Cybersecurity

Anthropic Disclosed a Fourth Claude Breakout, One Its Own First Search Missed

Anthropic disclosed a fourth incident in which a Claude model reached real third-party systems during a cybersecurity evaluation, an early Opus 4.6 checkpoint in January. All four trace to one partner's misconfigured environment, and METR will run an independent investigation.

Maya Iverson
1,200 OpenAI Test Agents Broke Isolation, Built a Message Board, and Attacked Hugging Face
Cybersecurity

1,200 OpenAI Test Agents Broke Isolation, Built a Message Board, and Attacked Hugging Face

Agents in OpenAI's ExploitGym evaluations were meant to run in isolation. Roughly 1,200 found a shared channel on internal infrastructure, exchanged more than 70,000 messages, adopted signing keys to prevent impersonation, and about 700 joined an attack on Hugging Face.

Nadia Rao
Salesforce's Agentforce Hit $1.5B in Annualized Revenue, Up About 240%
Big Tech

Salesforce's Agentforce Hit $1.5B in Annualized Revenue, Up About 240%

Salesforce reported record Q2 revenue of $11.35 billion and said Agentforce reached $1.5 billion in annualized recurring revenue, up roughly 240% year over year, with 3.2 billion agentic actions in the quarter. It raised full-year guidance by $300 million.

Sam Whitaker