Latest cybersecurity reporting from selected sources.

Also: AscendEX Shuts Down; Ethereum Foundation on AI Bug HuntingThis week, Ostium paused trading, AscendEX shut down, Ethereum said AI hastens bug hunting, not verification, the United States charged an inmate over seized transfer, ex-Sheriff's deputy jailed for extortion, Tangem card security, Iran-linked wallets sanctions and the U.S. may drop BitClub charges.

When combined with another exploit, the "PromptFiction" vulnerability, which has been fixed, could have enabled an end-to-end attack on a targeted system.

Old Unix Symlink Trick Lets Malicious Code Bypass User ChecksWiz researchers found that six popular AI coding assistants can be tricked into modifying sensitive files, including SSH keys, while their approval prompts display a harmless filename. The GhostApproval technique exploits a decades-old Unix symlink behavior to mislead users.

Researchers at the AI Now Institute developed a proof-of-concept exploit showing common AI tools used for security could backfire

Expect more compressed patching cycles from Apple going forward, as attackers leverage artificial intelligence to reduce time to exploit.

Security firm Sysdig says it has found what it believes is the first ransomware attack run from start to finish by an AI agent

"Agentjacking" is the latest demonstration of how easily attackers can exploit an AI agent's inability to differentiate between content and instructions.

Threat actors are continuing to exploit a critical Langflow vulnerability as part of fresh attacks designed to deliver a Monero cryptocurrency miner

You don't need Mythos or GPT-5.5-Cyber to find a vuln to exploit when the world's password habits are so sloppy

Four vulnerabilities allow attackers to exploit Dify, a platform for AI application building and management, to silently access and exfiltrate sensitive data.

Microsoft researchers have detailed an exploit chain, named AutoJack, that turns an AI browsing agent into a delivery vehicle for remote code execution

AutoJack is a novel exploit chain showing how a single malicious webpage can turn an AI browsing agent into a remote code execution vector on the host machine. By abusing trust in localhost, missing authentication, and unsafe parameter handling, attackers can trigger arbitrary process execution through AutoGen Studio’s MCP WebSocket. The research highlights a broader pattern - when agents can browse untrusted content and access local services, traditional boundaries like localhost are no longer secure. The post AutoJack: How a single page can RCE the host running your AI agent  appeared first on Microsoft Security Blog.

Athena is a new an industry coalition to fix the vulnerabilities frontier AI models find before attackers can exploit them

Stuff broke again. Not in a movie way. An old tool was left exposed. An abandoned package was abused. A deprecated feature was still running in prod

Lightspeed Funds Will Support Defenses Against Continuous, Machine-Led ExploitationA Security, founded by former Sygnia executive Yossi Torati, emerged from stealth with $37 million to build defenses against weaponized AI that can automate discovery, exploit attack paths and manipulate agentic systems faster than human security teams can respond.

Evaluations of Claude Mythos 5 Elevates Offensive Cyber, But Isn't Fully AutonomousAnthropic says its new Claude Mythos 5 model that debuted Tuesday can consistently discover vulnerabilities, build exploit chains and assist attacks on weak enterprise networks, but remains below the threshold for fully autonomous large-scale cyber operations.

Anthropic's Mythos Preview was highly effective at finding vulnerability candidates, especially when analyzing source code. XBOW explores how the model performed across exploit discovery, reverse engineering, and live-site validation. [...]

MultiCare Health CISO Jason Elrod on Need for Faster Cyber ResilienceEmerging AI tools can identify and exploit software vulnerabilities within minutes, forcing healthcare organizations to rethink cyber strategies. Jason Elrod, CISO of MultiCare Health System, explains why exploitability management, microsegmentation and AI-driven resilience matter more than ever.

Also: Former Hodlnaut CEO Charged and Stake DAO Hit by ExploitEvery week, ISMG rounds up cybersecurity incidents in digital assets. This week, the U.S. sanctioned Iran's largest exchange, ex-Hodlnaut CEO faced charges, the U.S. Securities and Exchange Commission sued over a $12.3M AI crypto scam and exploits hit Gravity Bridge, Stake DAO and Gnosis Pay.

A Bugcrowd researcher has unveiled ExploitBench, an independent benchmark of AI models for vulnerability exploitation