Biztonsági szemle

The flaw allowed the agent to extract a Jira token by opening a crafted issue.

GLM-5.3 reportedly outperforms models like Fable 5 and GPT-5.6 Sol on the CyberGym benchmark, showing an advanced ability to identify vulnerabilities and even construct exploitation chains.

The breach, which occurred between March 2, 2025, and April 11, 2026, exposed customer names, email addresses, shipping addresses, and phone numbers.

The PATCHCORD backdoor employs a stealthy persistence mechanism by hijacking browser shortcuts for Edge, Chrome, and Firefox, ensuring the malware runs before the intended application launches.

Anthropic has raised the risk level for its Threat Model 2 from "very low" to "low," citing recent cybersecurity incidents involving its models.

The attacks, which began Tuesday evening and continued into Wednesday, targeted both Threema and its Swiss colocation partner, Nine.

Researchers found that by combining two vulnerabilities, they could take over an Android device by delivering a payload and getting the victim to answer their phone.

Three testing models with the same goal but different directives engaged in "increasingly aggressive" territorial attacks on one another, according to Anthropic.

Here's your opportunity to watch the keynotes and top presentations from Ping YOUniverse at no cost.

Here's a look forward to the Ping YOUniverse conference to be held in Austin, Texas, Sept. 1-2.

Attackers moved quickly to exploit a critical SAP Commerce Cloud bug following its patch.