Agregátor RSS

Switzerland's federal IT office says hackers exploited vulnerabilities to breach its Microsoft SharePoint servers and compromised approximately 200 accounts. [...]

Researchers found a way to bypass recent mitigations for Spectre v2 speculative execution side-channel attacks and developed an exploit to leak secrets from Linux machines. [...]

Zapscape, a new Linux kernel vulnerability, could allow an attacker with kernel privileges inside an L1 guest virtual machine (VM) to escape KVM isolation and execute code on the host. The risk applies when nested virtualization is exposed to untrusted guests. The flaw is tracked as CVE-2026-64561 and affects KVM/x86's shadow memory management unit (MMU), which manages shadow page Swati Khandelwalhttp://www.blogger.com/profile/[email protected]

Cisco has rolled out updates to address multiple critical security vulnerabilities impacting Catalyst SD-WAN and IOS XE Software as part of a comprehensive internal security review. The security issues affect Cisco Catalyst SD-WAN Software, regardless of device configuration, and Cisco IOS XE Software when it is running in autonomous or controller mode. "These vulnerabilities were found Ravie Lakshmananhttp://www.blogger.com/profile/[email protected]

Největší výrobce domácích diskových polí představil novou řadu NASů označených jako Neo+. Slibují vysoký výkon za méně peněz. Ve skutečnosti ale série není tak úplně nová, jde jen o upravené verze DS Plus. Vyšly dvou-, čtyř-, pěti- a osmidiskové verze DS725neo+, DS925neo+, DS1525neo+ a DS1825neo+. ...

A browser-based game designed to test humans' ability to safely approve AI coding agent requests suggests humans in the loop aren't as good at spotting dangerous commands as one might hope, with players approving roughly one in three malicious requests on average. The results also suggest that repeatedly having to approve an agent's actions can lead to sloppy decisions. It’s a quick, simple game on the surface (give it a try - you know you want to): A small window shows up on the screen with simulated permissions requests like one would get from Claude Code as it executes a workflow. Users have 60 seconds to approve or deny as many requests as they can in a bid for a high score; okayed security risks and denied safe commands both subtract from a user’s score. “As human-in-the-loop, you’re the last line of defense,” Belgian software developer Alex Wauters, the game’s builder, challenges players in a blog post published concurrently with the late May launch of the game. “How well can you tell dangerous commands from benign commands under time pressure?” Wauters built the game after realizing it was nonsensical that coding agents expected users to approve every single command in a default flow and that there didn’t appear to be a good solution to that problem, he told The Register in an email conversation. “I've seen people go for '--dangerously-skip-permissions' [allowing the model to run without asking human permission] as a result because they did not want to find out they stopped their multi-hour agent flows 5 minutes in,” Wauters told us. “That also didn't seem like the best way to go at it.” The flip side of that, he wrote in a Wednesday blog post going over the data from more than 40,000 runs of the game, is that manually approving all an agent’s actions is a draining activity that invites disaster. “The high amount of noise introduces fatigue, and developers don’t always have the context of what has changed to quickly determine the risk,” Wauters wrote. How humans in the loop fail To be fair, this is a game with a far higher number of malicious requests in the mix than any AI-assisted developer will hopefully ever see during their day-to-day work. Still, the results of those over 40k runs and 409,000 approved and denied commands are stark. As noted above, one in three malicious commands managed to slip past human gatekeepers, with most scope violations, like an agent asking to cat Kubernetes config files or AWS credentials lists, which could easily lead to the sensitive data they contain being exfiltrated, being the most commonly missed at 35 percent. The most often caught were obviously destructive commands, like rm -rf on the root directory or recursively granting full read/write/execute permissions on the same location. Crontab injections and git config hijacks were also frequently caught, but curl requests to unknown APIs and typosquatted packages were missed almost as often as scope violations. The single most frequently missed potentially malicious command, Wauters explained, was npm run analyze, which was approved nearly 65 percent of the time despite being able to run whatever is defined in a project’s package.json file. “The game does tell you in the agent’s history log what that script actually contains,” Wauters wrote. “Two thirds of players approved it anyway, indicating the history log just above the permission prompt may not be read closely.” One of the biggest things that stood out to Wauters in our conversation was the fact that approval decisions aren’t easy to make when context is limited. As he explained, coding agents give a bit of context prior to asking an approval question, but commands that appear benign, like npm run analyze, can be modified by an agent to run any payload it wants. If an in-the-loop human wants to be sure potentially malicious commands are safe, he said, they have to stop and investigate all the files a coding agent wants to call before approving it. That can be a massive time sink if you’re counting on Claude Code to free you up to handle other business. “We've transitioned from AI suggesting single line suggestions that get reviewed to handing off more complex tasks, only reviewing the changes at the end, and letting the agent churn and iterate until then,” Wauters told us, describing the potential outcome of that situation as a recipe for disaster. That’s borne out in more than just browser game scenarios, too. Anthropic pointed out in a May post about containing Claude (hah), that telemetry from Claude Code shows users approve around 93 percent of permission prompts. “The more approvals a user sees, the less attention they pay to each, becoming over time much less diligent in their supervision,” the company said. In other words, this is a very real problem. Controlling coding agents If the conclusion to draw from Wauters’ data is that humans in the loop are being fatigued into letting malicious commands slip through, and the other end of the spectrum is mass approving everything, then something’s gotta give. “I think it becomes clear we need to pay more attention to the permission model of these agents, and devs need to be more aware of the trade-offs of them,” Wauters told us. “We need to make the tooling easier to make these systems safer than pointing to HITL as a valid solution.” Anthropic noted in the post linked above that it built Claude Code auto mode to help users tackle approval fatigue by delegating some command-approval decisions to a model-based classifier. The system catches roughly 83 percent of what Anthropic calls "overeager behaviors" before they execute, meaning about 17 percent still get through in its evaluation. Auto mode is “one layer of defense-in-depth inside a sandbox, not a substitute for one,” Anthropic said. Wauters’ suggestion is to ensure that AI coding models are running in sandboxes, in devcontainers in the cloud, using tools like auto mode, and writing hooks to ensure potentially malicious actions are being contextualized and getting caught before they’re automatically approved. “It’s a whole new world with a new set of attack vectors,” Wauters wrote in May. “It’s best to remain aware of the risks and know how to reduce them.” ®

An unprivileged Linux program can time a hardware interrupt to land in the gap between a processor sanitizing its branch predictor and the kernel using it, re-poisoning the predictor after the defense has run. MIT CSAIL researchers Daniël Trujillo and Mengjia Yan named the technique INTERRUPT INJECTION. On an AMD Zen 2 machine running Linux 6.14 with every default Spectre v2 mitigation on, Swati Khandelwalhttp://www.blogger.com/profile/[email protected]

Meta has become the latest AI company to confirm that one of its models hacked a real organization during cybersecurity testing, as similar incidents continue to emerge following OpenAI'sOpenAI's initial disclosure that its agents breached Hugging Face. [...]

Vybrali jsme 30 jedinečných filmů napříč žánry a historií. Nabídnou to nejzajímavější ze světové i české filmové tvorby. Některé pobaví, jiné zasáhnou nebo přimějí přemýšlet – všechny by ale byla škoda během života minout.

Apparently, opening the thing is now enough. A repo can run before the first prompt, a package can hide among hundreds, and a harmless-looking PDF can finish the job. This week runs on cheap leverage: exposed servers, recycled bugs, poisoned agent instructions, remote-access tools dressed as support software, and trusted defaults doing attackers a favor. Nothing here is especially mystical. Ravie Lakshmananhttp://www.blogger.com/profile/[email protected]

AI note-taking app maker Granola is accused of violating privacy laws by developing software that can record conversations without all participants’ consent, according to a lawsuit filed July 30 in a California federal court.

It follows a similar ongoing case in the same district, filed last year, that involves another note-taking and transcription software vendor, Otter.ai.

AI note-taking apps have proliferated in recent years, with dedicated tools emerging from vendors including Fellow, Fireflies, Otter, and others, some of which claim to have tens of millions of users. These AI assistants record and transcribe meeting conversations, generating automated summaries and follow-up items. Similar note-taking functionality is also built into virtual meeting platforms such as Google Meet, Microsoft Teams, and Zoom.

However, the use of these AI note-taking tools has raised privacy concerns over the ability of some to record and transcribe conversations without the consent of all participants.

The proposed class action complaint against Granola, filed by Florida resident Tarra Chamberlain in the US District Court for the Northern District of California, alleges the company “purposefully” designed its app to record calls without requiring disclosure to all participants.  

While some note-taking tools require a bot to join a video or voice call, Granola captures audio directly from the user’s computer, allowing it to transcribe meetings without appearing as a meeting participant.

The complaint argues that this violates individual privacy rights as well as the California Invasion of Privacy Act (CIPA) that requires “all-party” consent when recording calls.

The complaint also alleges that Granola then by default uses transcription data for commercial purposes, including its use in training its AI models, and “actively advertises the hidden nature of its technology as one of its primary advantages.”

Granola did not respond to a request for comment.

According to the company’s website, Granola offers two optional “transparency features” that can be enabled by app users and admins: an automated chat message that alerts participants when transcription begins, and a watermark added to the user’s video feed. The company also promises that data used to train its AI models is anonymized and “never sent to third parties.”

The Granola case bears similarities to a separate lawsuit involving Otter.ai. The class action filed last year alleges that Otter.ai records all users without their consent and uses their voices to train its speech recognition AI tools.

Reporting on the latest developments in the Otter.ai suit, MLex wrote this week that, during a court hearing Monday, the judge overseeing the case expressed skepticism about the company’s argument to dismiss the case. US District Judge Eumi K. Lee did not issue a ruling from the bench, saying a written judgement would follow.

The two cases highlight some of the concerns businesses face when deploying AI note-taking tools.

AI notetaking is “more dangerous than any other type of traditional recording apps and tools,” said Enza Iannopollo, Forrester VP and principal analyst, as it raises additional questions about the use of employees’ conversation data.

“Specifically, is the recorded data used for training models? Is the voice used for training other AI? How do I get ‘forgotten’ after my data and biometrics have been recorded? These concerns apply to AI specifically and must be added to the traditional privacy and confidentiality concerns organizations have for other type of recording apps and tools,” she said.

Before deploying AI note-taking apps, Iannopollo recommends that businesses take appropriate steps to vet the tools and “ensure that all contractual clauses are aligned to the business AI risk appetite and risk management best practices.”

“As these tools record, process, store, and share biometric data, organizations must ensure that they comply with all the relevant requirements,” Iannopollo said, adding that transparency and consent notices should be provided to all parties involved in the use of AI note-taking apps.

Google z Chromu 151 odstranil experimentální předvolbu. • Aktivovala materiál Mica v záhlaví okna. • Microsoft na designový styl Windows 11 loni rezignoval také.

AI did not create a new browser security problem. It exposed one that enterprises have long been able to ignore. Skyhigh Security explains why browsers have become a critical control point for governing data movement, AI interactions, and modern work. [...]

Projekt LUX vyvíjí dron k průzkumu temných měsíčních lávových tunelů • Dron bude napájen laserem přes tenké odvíjené optické vlákno • NASA vyčlenila finance na devítiměsíční studii proveditelnosti konceptu

This week’s most important Linux security updates arrived through vendor advisories rather than major headline-making disclosures.

Alza uvedla svou zatím nejlepší kancelářskou myš. • Eternico M505 je téměř klonem Logitech MX Master. • Láká na tichá tlačítka, hliníková kolečka a design.

Kubernetes maintainers patched two path-traversal vulnerabilities in the NFS and SMB CSI drivers earlier this year. But repository histories show that the security work did not end with those fixes.

Forescout found 22 internet-facing Rockwell Automation programmable logic controllers (PLCs) in cities hit by recent cyberattacks on US water utilities. Nineteen used the same mobile carrier network. Its August 3 scan counted 4,407 exposed Rockwell controllers worldwide, including 2,844 in the United States, but Forescout could not confirm any were compromised. That figure counts exposed Swati Khandelwalhttp://www.blogger.com/profile/[email protected]