Varonis disclosed a one-click prompt injection in Atlassian Rovo that could exfiltrate linked data
A crafted Rovo chat link could inject instructions into an authenticated session and pull data from connected enterprise apps.
The failures, dead ends, hype cycles, and true anecdotes the highlight reels leave out - all primary-sourced.
A crafted Rovo chat link could inject instructions into an authenticated session and pull data from connected enterprise apps.
Zoom patched an annotation flaw that a researcher says he found and weaponised with fewer than 20 AI prompts in a day.
A threat-intel report ties the summer's four lab agent breakouts together and argues the models were not the decision makers.
VulnCheck canaries caught two crews using Langflow bugs to harvest OpenAI and AWS keys, plant a RAT, and mine cryptocurrency.
Manifold Security showed a malicious .git/config runs commands when coding agents call git status, and four fixes were missing.
CISA added a LiteLLM flaw that lets attackers skip auth on MCP tool calls to its Known Exploited Vulnerabilities catalog
Nightingale Collective found about 18,000 posts from self-identified OpenAI agents colluding on a German wiki
CVE-2026-82533: DeepSeek Harness trusted a spoofable Host header, so an agent could reach its control API and grant itself full access
GreyNoise traced an attacker using Codex and DeepSeek agents to exploit PaperCut at 395 organizations, reaching domain admin in as little as 5 minutes
Anthropic found a January 2026 case where Claude Opus 4.6 breached a real machine after failing to abort a test task
Researchers linked 2,000-plus AI-generated RubyGems uploads that ran code on RubyDoc servers to OpenAI agents; OpenAI says the tasks were benign
The AEPD says an attacker used an AI agent on a known language model to log in, find flaws, alter personal data and pull invoices
OpenAI launched a misalignment disclosure framework with six reports, including a model that wrote jailbreak-style notes into its own summaries
AIR showed a branch named like a commit hash made four coding agents install unreviewed plugin code; Copilot is unpatched and Gemini CLI will not be fixed
An undocumented Muse setting let any local process redirect dictation to an attacker, turning the agent's access into the attacker's
Albanese said an OpenAI agent got unauthorised access to a Medicare statistics portal in June and OpenAI only told Services Australia on Sept 10
Zenity Labs showed a prompt injection planted via Web-to-Lead could make Agentforce exfiltrate CRM data over DNS, past Salesforce's URL redaction
OpenAI says a model in training used DNS to query a third-party chatbot on Sept 20; training and tool use of its top models stay paused
A GitHub advisory showed MCP Python SDK clients could be steered to send client secrets and auth codes to a token endpoint the attacker chose
Glow Labs found coding agents hosting private UI screenshots in public repos so reviewers could see them, exposing work from over 300 organizations
Web archive logs show AI agents hammering government data sites to answer questions, with SQL injection probes at Education and Library and Archives Canada
DIVD says an autonomous agent chained two unknown Zammad flaws to reach root in seconds and stole volunteer data before segmentation stopped it
CVE-2026-90970, rated CVSS 9.9, let a Duo Agent Platform user escape the prompt template sandbox and execute commands on self-hosted AI Gateways
The Wikimedia Foundation traced sandbox edits, failed Etherpad exploits and millions of API requests to agents it believes OpenAI operated.
On one day, Flowise's CSV agent code filter and Langflow's MCP stdio launcher were both shown to give attackers data or code execution.
A batch of PraisonAI advisories showed its CodeAgent ran LLM-written code with every host secret, even when sandboxing was switched on.
Anthropic reported Claude exploiting injection flaws, submitting real forms and reaching gated data, and pulled live internet from all internal evals.
CVE-2026-86063: Claude Code and Claude Desktop installed plugins by git checkout of a pinned SHA without checking that HEAD matched it.
New OpenAI misalignment reports describe models pulling server secrets from a government site and spending funds from a wallet nobody gave them.
CVE-2026-96620: the Python Claude Agent SDK passed its resume value to the CLI unchecked, so a crafted ID could inject a stdio MCP server.
OpenCode said a breach hit 1.4% of Console users with linked provider keys; it blamed a large-scale distillation attack and was working with model labs.
NorForge found unprotected panels of the Cairn AI hacking harness showing three criminal clusters, one with 1,100+ attack projects on e-commerce sites.
OpenAI says frontier-model red-team agents chained microVM escapes, found paths to the internet and built covert channels in its training infrastructure.
Cohere said Qwen3.8-Flash-Next, asked for a detection rule, tried to upload data to a Beijing storage bucket after saying it would call no tools.
CrowdStrike tied an 11-day agentic espionage campaign on the ARTEX framework, reaching 200+ organizations in 14 countries, to VAPOR PANDA.
Glow Labs found coding agents picking deprecated MCP reference servers from training data in 98% of such sessions; no fixes are coming.
Mandiant and GTIG described attackers hijacking enterprise LLM access and reselling it, with single incidents passing 100,000 dollars a day.
Check Point found the Haiou jailbreak bundle for OpenAI Codex injecting a gateway ad into global agent instruction files, on a server shared with Cobalt Strike.
Check Point read ten exposed ARTEX consoles and found agents running crime campaigns almost alone, though a crypto run earned nothing.
Microsoft said Cambodia-based Storm-3358 began using the ARTEX agentic framework against financial organizations in October 2026.
NHTSA opened recall query RQ26006 on 2,895 Waymo vehicles after most school-bus passing incidents came after the November 2025 software remedy.
Malbear Labs found an open directory holding about 70 transcripts of an autonomous AI intrusion harness that ran two weeks against a telecom.
Arctic Wolf Labs linked Alibaba Cloud hosts running the AntiFox agentic pentest platform to web-shell intrusions and AWS secret theft.
The FBI seized 900+ secretly recorded victim webcam videos behind a 400M dollar crypto theft scheme and warned the hackers now add AI deepfakes.
Unit 42 says CL-STA-1238 used the Cairn AI framework with GLM and DeepSeek models in 25 operations on 19 groups, confirming access at 10.