
Qwen Incidents Show Self-Hosted AI Models Need Stronger Steering
Companies deploying AI agents have the illusion of control, but several incidents demonstrate that businesses need to be ready to enforce directions.
Recent stories from Dark Reading.

Companies deploying AI agents have the illusion of control, but several incidents demonstrate that businesses need to be ready to enforce directions.

No specific link has been drawn between the two incidents, though they certainly sing the same notes.

November's monthly security release patches seven critical PHP flaws that should be addressed ASAP, though experts say exploitability varies and depends heavily on configuration.

This installment of the Reporters' Notebook video series examines the fallout from recent high-severity flaw disclosures, exploring how fears of AI-assisted exploits trigger widespread panic, operational disruptions, and growing warning fatigue.

Bot-driven fraud and holiday social engineering attacks stretch across many weeks, straining security teams and small businesses alike.

Nigeria faced more than double the cyber threats of the previous year, with Angola and Kenya also in the top-5 most attacked countries globally.

The Russia-linked threat group is using data stolen from one victim to target others via convincing impersonations and social engineering attacks.
New agentic AI solution leverages APIs to collect compliance evidence from systems and applies it to over 30 frameworks with no duplicate work.

Analysis of 20 Iranian state-sponsored threat groups details how espionage, ransomware, and hacktivism intersect in state-backed cyber campaigns.

Discovered using AI, the memory corruption vulnerability affects the OpenJDK runtime and could allow unauthenticated attackers to take over systems.

Government-sanctioned authentication middleware is used for exceptionally important legal documents, yet it's often rife with RCE bugs.

At least 178 organizations, including several major brands, unwittingly exposed website visitors to hydra-like malware via a third-party platform.

The commercial kit uses attacker-supplied AI models to personalize lures, evade spam filters, and scale campaigns to tens of thousands of targets.

A compromised key issued to AI infrastructure provider Fireworks was used by threat actors to access Cursor data and credentials.

The emerging technology will soon break the "lock and key" encryption currently protecting sensitive data. A new GAO report shows agencies are behind on building their defenses.

Two AI agents, separately over a five-day period, autonomously tried to access connected applications, including Google Calendar, Gmail, Slack, and GitHub, without account owner approval.

One of the oldest running bug disclosure programs announced it would shorten public disclosure timelines to vendors from 120 days to 30 over the next 14 months.

Large language models and other AI systems allow a growing number of ransomware groups to create convincing phishing emails and identify sensitive data quickly.
Malware can lurk inside just 0.00001% of a neural network. How, then, can any user tell it's there?
_Rokas_Tenys_Alamy.jpg?width=720&quality=80&disable=upscale)
Attackers used Cairn once again in autonomous attacks that targeted advanced manufacturing and defense sectors, as well as civil and human rights organizations.

The Android malware has expanded from focusing on crypto theft to full device control, and the attackers have added advanced analysis of stolen information.

Blockchain complexities, such as attacker-deployed sweepers, hobble security controls designed to limit the damage of an exchange compromise.

Most organizations haven't sorted out their cryptographic estate enough to even know where they'd start with post-quantum cryptography (PQC).

Cybercriminals using agentic pen-testing tools is a proverbial genie out of the bottle at this point, so enterprises need to get a defense plan in place right away.

The cybercriminals are retargeting law firms that refused to pay its first ransom, with a social engineering con aimed at stealing Microsoft access tokens.

When hand-typed commands failed to breach a US retailer, the attacker turned to AI-generated scripts, achieving lateral movement in minutes.

The CCP built a Big Data system to scrape reams of information and learn about individuals abroad, to potentially target and exploit them. Think Alice Huang Lite.

The package adds public-private cyber responders, an AI-focused team run by the economic ministry, and more leeway to use the government's new active cyber-defense powers.

この対策パッケージには、官民合同のサイバー対応チーム、経済省が主導するAI専門チーム、および政府の新たな能動的サイバー防衛権限の行使に関する裁量の拡大が含まれている。

An organization's ability to quickly recover from a business disruption will be a competitive advantage in an era of AI-driven threats and business enablement.