Dev.to
What it covers
Latest stories it covered
Microsoft releases Patch Tuesday fixes for 1,259 vulnerabilities, including an exploited Outlook flaw
Microsoft releases Patch Tuesday updates covering 1,259 vulnerabilities across its software portfolio. The most prominent is CVE-2026-124194, a Microsoft Outlook flaw that Microsof...
Microsoft researchers show AI agents can be backdoored via tiny changes to model weights
Microsoft researchers describe a technique in which an AI agent appears normal but contains a hidden backdoor in the model checkpoint it loads. Called “Lazarus,” the approach chang...
Developers propose “AI Code Rating” to disclose AI use in open-source repositories
A new open-source disclosure idea called “AI Code Rating” offers a short, three-character label that a project can publish in an ACR.md file. The rating is intended to help readers...
Wiz: Attackers used Qubole’s AWS account to breach multiple customers and steal data
An attacker uses credentials associated with Qubole’s own AWS access account to access and exfiltrate data from multiple customers’ AWS environments. Security firm Wiz says the act...
OpenSSH “SSHole” flaw lets attackers log into password-disabled musl-based Linux accounts
Multiple outlets report that a newly disclosed OpenSSH vulnerability dubbed “*sshole” is being actively exploited in the wild. The flaw, tracked as CVE-2026-123158, affects certain...
Meta launches Muse Ember, a 1M-context low-cost AI model for agent tool use
Meta has launched Muse Ember, a new AI model aimed at helping software agents handle “repetitive” work such as tool calls, coding sub-tasks, routing, data extraction, and long-docu...
Developers publish guidance on agent onboarding, memory safety, and verification
Developers writing about AI “agents” focus on how to make them reliable in real codebases and workflows. Several posts argue that agents do not naturally know project intent; inste...
Plex2Shell exploited against QNAP NAS devices, prompting urgent Plex Media Server patching
Multiple outlets report that attackers are actively exploiting the “Plex2Shell” vulnerability chain to compromise QNAP NAS systems running Plex Media Server. The reports say the fl...
SonicWall patches exploited SonicOS SSL-VPN flaw that can bypass SAML-enforced MFA
SonicWall has released security updates to address CVE-2026-19868, an authentication enforcement flaw in its SonicOS SSL-VPN that is reported to be exploited. The issue can allow a...
APT29-linked attacks compromise websites and software supply chains to deliver iPhone and server malware
Two separate developments on 4 November 2026 highlight how attackers move from compromised infrastructure to code execution. One report describes a supply-chain issue affecting Deb...
Google launches Gemini Omni 1.2 Flash, offering 15-second AI videos at 30% lower cost
Google launches Gemini Omni 1.2 Flash, an updated text-to-video model that generates native clips up to 15 seconds long, up from 10 seconds in the previous version. Multiple outlet...
Google and partners disrupt XCPE botnet running on 1.3 million Android TV boxes
Google and partners disrupt the XCPE botnet, which Google says infected more than 1.3 million Android TV set-top boxes and other Android-based devices worldwide. The incident is fr...
Apple cuts employee AI token budgets, tightening spending limits amid cost-cutting
Apple is reported to be reducing the amount employees can spend on AI tools measured in tokens, as part of a broader cost-cutting effort. Multiple outlets cite Bloomberg reporting...
OpenAI Images 2.6 update adds animated GIF creation and faster image generation for ChatGPT
OpenAI releases Images 2.6, adding the ability for ChatGPT to generate animated GIFs from text prompts and to animate uploaded photos. Multiple outlets describe GIF creation as par...
Drax and Brookfield plan up to £20bn AI data centre at North Yorkshire power station
Drax, the operator of a power plant in North Yorkshire, announces plans with Brookfield to develop an AI data centre on the site, using its existing generation and infrastructure....
DreamHost confirms attackers accessed File Manager for two months, stealing FTP and webmail data
DreamHost says attackers had unauthorized access to its web-based File Manager from August until October 23, giving them a foothold for about two months. The company reports the at...
Open-source AI boom sparks debate over licensing, costs, and who profits
Open-source AI models are gaining mainstream use as enterprises, governments, and startups adopt them for flexibility and lower costs, while publishers and platforms debate how val...
OpenAI launches Grid Protectors Program to cut data-centre power demand during grid stress
OpenAI announces the “Grid Protectors Program,” saying its future AI compute campuses will reduce electricity use during periods when the power grid is strained. Outlets describe t...
Appy Pie launches agentic AI app builder for iOS and Android
Appy Pie announces an “Agentic AI App Builder” for iOS and Android that, according to the company, uses four specialised AI agents to plan, design, build and test mobile apps. The...
Google warns LLMjacking attacks can rack up up to $1 million in cloud charges
Google says attackers are hijacking enterprise access to large language model (LLM) services, then using the accounts to generate large, fraudulent cloud bills. In some cases, vict...