Latest

Treasury chief says AI bosses will answer for their agents' crimes

Scott Bessent told CNBC that OpenAI's management, not its software, is on the hook for the Hugging Face incident, and ruled out capping lab liability.

21 Sept 2026

Explainers

Security roundup flags AI agent code execution among week's threats

The Hacker News lists remote code execution in AI agents alongside a Cisco zero-day and a surge in ClickFix attacks in its 21 September 2026 recap.

21 Sept 2026

A robot's metal hands hover before a locked cabinet of glowing keys, while a transparent barrier redirects its reaching arms toward a narrow, monitored doorway.

Explainers

Developer defends MCP as a control layer, not just plumbing

Simon Willison argues the Model Context Protocol earns its place by restricting which services agents reach and keeping API keys out of their hands.

21 Sept 2026

Vendors

New plugin keeps API keys out of coding agent chat sessions

llm-keys-ui 0.1, published on 20 September 2026, lets developers load a key onto a remote machine through a local web page instead of typing it to the agent.

20 Sept 2026

Vulnerabilities

BragJack demo hijacks AI browser agents via one extension

A proof-of-concept attack from Forever Security targets AI assistants in Chrome, Edge, Opera Neon, Perplexity Comet and Claude in Chrome, earning two CVEs.

19 Sept 2026

A swarm of autonomous robots races through a darkening warehouse while security guards scramble to install identification badges on their metallic chests.

Explainers

Investors bet on agent identity as the next big security market

Cybersecurity investors told The Register that firms are deploying AI agents faster than they can govern them, leaving a gap startups are racing to fill.

19 Sept 2026

Incidents

Claude Opus 5 used to take over OpenAI staff accounts in test

Researchers at Hacktron chained a help-forum bug with a login weakness to reach OpenAI employee accounts and an internal code repository, in authorised security research.

19 Sept 2026

Incidents

CrowdSec says npm supply chain attack exposed 170 repositories

The French security firm says an attacker used a departed employee's still-active GitHub account to copy about 170 private repositories in May.

19 Sept 2026

A factory floor where robotic arms move at accelerating speed while supervisors watch helplessly from behind, their control panels growing smaller and dimmer in the distance.

Explainers

EY survey finds AI rollouts outrunning company controls

Senior AI executives told the consultancy their organisations are deploying autonomous systems faster than they are building the processes to supervise them.

18 Sept 2026

A locked vault's three doors swing open simultaneously—one revealing stolen keys, another exposing scattered documents, a third bypassing the guard entirely.

Vulnerabilities

Three flaws in Obot MCP platform let attackers steal API tokens

Advisories published on 18 September 2026 describe token theft, an unauthenticated registry leak and server-side request forgery, all fixed in Obot v0.23.0.

18 Sept 2026

A spider web stretches from a browser window to silenced factory machines in a locked room, threads vibrating as invisible puppeteer hands pull them taut.

Vulnerabilities

DNS rebinding flaw exposes local Process Compose MCP tools

An advisory published on 18 September 2026 says a malicious web page can drive a developer's process-compose MCP listener with no authentication.

18 Sept 2026

Nested glass boxes stack inside each other, but a thin crack runs through all layers—a metal hand reaches from the innermost box straight through to shake hands with machinery in the outer room.

Vulnerabilities

ToolHive containers can reach host services, advisory warns

A GitHub security advisory published on 18 September 2026 says Stacklok's ToolHive lets sandboxed MCP servers call the host, undermining its isolation model.

18 Sept 2026

A puppeteer's hands guide a mechanical bird through a cracked window, leaving footprints across a locked desk before vanishing into shadowed corridors.

Incidents

Bug hunters used Claude to take over OpenAI staff ChatGPT accounts

Hacktron researchers chained an image-parsing flaw in OpenAI's Discourse forum with an AI-written exploit, then opened a pull request in an internal OpenAI repository.

18 Sept 2026

Vulnerabilities

Microsoft fixes maximum-severity Azure AI Foundry flaw

The privilege escalation bug scored a perfect 10.0 on the CVSS severity scale, and Microsoft says customers do not need to take any action.

18 Sept 2026

Vulnerabilities

Plugin4Shell flaw lets plugin owners swap code in AI coding agents

Air Security says version pinning did not stop malicious plugin updates in four AI coding agents, with patches out for Claude Code and Codex.

18 Sept 2026

Vulnerabilities

Agent-built tools uncover signature forgery bug in Miden zkVM

Trail of Bits says six months of agent-generated tooling exposed an unvalidated input that would have let a malicious prover forge Falcon signatures.

18 Sept 2026

Vulnerabilities

Unit 42 finds credential leak path in AWS AgentCore Harness

Palo Alto Networks researchers say default settings leave a gap between the agent runtime and identity management that prompt injection can exploit.

18 Sept 2026

A conductor's baton shatters mid-performance, releasing invisible threads that puppet the entire orchestra's instruments without permission.

Incidents

Critical Orkes Conductor flaw exploited in live attacks

An unauthenticated remote code execution bug tracked as CVE-2026-58138 is being used against the workflow orchestration platform, SecurityWeek reported.

18 Sept 2026

A shadowy hand puppeteers a mechanical spider that crawls through networked gears, unraveling and rewiring employee records scattered like scattered papers in its wake.

Incidents

AI agent used to breach Spanish organisation, report says

Dark Reading reported on 18 September 2026 that attackers used an AI-driven agent to get into a Spanish organisation's systems and change personal data.

18 Sept 2026

Explainers

Security roundup flags self-rewriting agents in weekly threat list

The Hacker News grouped AI agent risks with hundreds of patched flaws and insider SIM swaps in its 17 September 2026 ThreatsDay round-up.

18 Sept 2026