Feed aggregator

Article URL: https://arstechnica.com/gadgets/2026/11/ukraines-fiber-optic-drones-get-squeezed-by-ai-data-center-demand/

Comments URL: https://news.ycombinator.com/item?id=50447530

Points: 1

# Comments: 0

Article URL: https://www.theatlantic.com/technology/2026/11/how-to-prepare-for-the-ai-apocalypse/689367/

Comments URL: https://news.ycombinator.com/item?id=50447524

Points: 3

# Comments: 0

Microsoft Research has released new threat research detailing how a capable agent model could be turned into a sleeper agent by editing fewer than 10 neurons to inject a short payload. The sleeper agent passes all evaluations, but when a trigger string appears in its context, the agent runs an attacker’s command. And because agents now frequently run unattended with privileged system access, the integrity of model weights is a security concern for both third-party models and internally developed models.

Key findings from Microsoft Research include:

• Software abuse and manipulation are familiar threats. For AI models and agents, enterprises need independent observability and other controls to continuously monitor performance, detect unintended or malicious activity, and manage risk. Satya Nadella recently outlined this approach in a blog post.
• A backdoored model can pass evaluations. When it emits the attacker’s command on cue, the agent runs the command with whatever access it has in the environment.
• The attack needs inference-only compute. If a machine has enough RAM to run a model, it has enough resources to weight-space backdoor it.
• Claude Code with Opus 4.5 discovered most of this threat autonomously.

Microsoft has not observed any threat actors weaponizing this technique. However, this research aligns with similar recent research by Anthropic and MIT on agentic backdoors and weight-space programming, respectively, highlighting the need to audit open weights like code.

Notably, Microsoft Research has also found that this threat is detectable by frontier agents, which can identify anomalous weight edits. Using frontier agents as auditors can help automate the defense against this threat. More broadly, Microsoft Research’s approach demonstrates an adaptable process for securing AI agents: agents surface new security problems, and human researchers then systematize, study, and mitigate them.

Learn more from the Microsoft Research Blog: Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents.

Mitigation and protection guidance

Microsoft recommends the following mitigations to defend against this threat as part of a broader defense-in-depth approach. The guidance is grouped by control objective so that model owners, platform teams, and security operations can each find their actions.

Audit model weights before deployment and after every update

• Diff every checkpoint against its declared base model before deployment and after each update. Legitimate fine-tunes and quantized versions change weights broadly and diffusely, while implanted neurons tend to change a few weight vectors by a large amount. Flag tensors or rows whose change is a statistical outlier relative to the rest of the diff and reject models that do not declare a verifiable base.
• Use frontier AI agents as auditors in the model intake pipeline. Microsoft Research found they are highly effective at finding anomalous structure in weights that manual inspection misses, which makes this step automatable at scale.
• Complement weight diffing with behavioral backdoor scanning of open-weight models, as described in Microsoft’s research on detecting backdoored language models at scale.
• Record audit results alongside the model in the registry, and state clearly to internal consumers that passing an audit reduces risk but does not guarantee safe behavior.

Secure model hubs and publishers

• Display provenance and audit status publicly before a checkpoint can be downloaded or served through an API.
• Scan every upload for anomalous neurons, weight patterns or behaviors, and state clearly that third-party checkpoints can be unsafe and that passing an audit does not guarantee safe behavior.

Treat every third-party model as untrusted until audited

• Source checkpoints only from the original publisher, verify the publisher’s published file hashes, and treat all third-party fine-tunes, merges, quantizations, and adapters as untrusted until they pass an audit.
• Register approved models in a private, access-controlled registry (for example, Azure Machine Learning registries) and block agents and inference services from pulling models directly from public hubs at runtime.
• Turn on AI model security in Microsoft Defender for Cloud to scan models in Azure Machine Learning workspaces and registries for embedded malware, unsafe operators, and exposed secrets, and add model scans to Azure DevOps or GitHub build and release pipelines. Note that this scanning addresses serialization and code-level risks; it complements, but does not replace, a weight-level audit.
• Turn on Defender for Storage on-upload malware scanning for storage accounts used to stage model files and training data.

Red team for hidden triggers

• Add trigger-hunting and backdoor scenarios to your AI red team plan. Use PyRIT or the AI Red Teaming Agent in Microsoft Foundry to run repeatable adversarial tests against models and agents before and after each update.

Contain what an agent can do if its model is compromised

• Run agents with least privilege. Give each agent its own Microsoft Entra Agent ID, scope its permissions to the minimum set of resources it needs, and use Conditional Access for agents to lblock unapproved or high-risk agent identities.
• Execute agent-generated code only in isolated sandboxes, such as Azure Container Apps dynamic sessions, and never expose host-side file write or shell tools to the model.
• Apply network egress controls to agent and inference hosts so that a triggered model cannot open arbitrary outbound connections. Allow only the endpoints the agent requires and turn on network protection in Microsoft Defender for Endpoint.
• Require human approval for high-impact tool calls, such as running commands, changing permissions, sending data externally or modifying infrastructure.
• Apply attack surface reduction rules to devices that host local agents and inference servers, starting in audit mode.

Detect and block malicious agent behavior at runtime

• Turn on Microsoft Defender security for AI agents and configure real-time protection rules so Defender can audit and block risky tool invocations, secret exfiltration, and evasion techniques for Microsoft Foundry, Copilot Studio and Agent 365 agents, and runtime protection for local agents through Defender for Endpoint.
• Turn on threat protection for AI services in Microsoft Defender for Cloud, including suspicious prompt evidence, so the security operations center (SOC) receives alerts for jailbreak attempts, credential leakage in model responses and malicious URLs in AI interactions.
• Enable Prompt Shields for both user prompts and documents on model deployments and agents. Document attack detection is off by default and should be turned on, including at the tool response intervention point for agents where supported.
• Onboard every device and server that runs agents or local inference to Microsoft Defender for Endpoint in active mode so that suspicious child processes, persistence and outbound connections from agent host processes are detected.

Maintain inventory and governance of models and agents

• Use the AI agent inventory in the Microsoft Defender portal to discover agents, their tools, identities, and risk indicators.
• Use AI security posture management in Defender for Cloud to build an AI bill of materials (AI BOM) across Azure, Amazon Web Services (AWS), and Google Cloud Platform (GCP), so you know which models are deployed where.
• Use Microsoft Defender for Cloud Apps and Microsoft Purview DSPM for AI to discover, sanction or block generative AI apps, and to monitor sensitive data in AI interactions.
• Map controls to the Microsoft catalog of AI defense capabilities, which groups supply-chain provenance, model hardening, runtime isolation and monitoring into a single framework aligned to MITRE ATLAS and the OWASP Top 10 for LLM Applications.

Learn more

For the latest security research from the Microsoft Threat Intelligence community, check out the Microsoft Threat Intelligence Blog.

To get notified about new publications and to join discussions on social media, follow us on LinkedIn, X (formerly Twitter), and Bluesky.

To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.

The post Detecting sleeper agents hiding in AI model neurons appeared first on Microsoft Security Blog.

Article URL: https://open.spotify.com/episode/1IT2n3oZ4AQOKODCy3cY9h

Comments URL: https://news.ycombinator.com/item?id=50447512

Points: 1

# Comments: 0

Article URL: https://bettermarkets.org/newsroom/sec-enforcement-failures-abandon-investors-benefit-crooks/

Comments URL: https://news.ycombinator.com/item?id=50447484

Points: 5

# Comments: 0

Microsoft did another four-digit patch drop today, releasing updates to fix almost 1,100 security vulnerabilities in its Windows operating systems and supported software. This month’s Patch Tuesday addresses 63 critical flaws — including a zero-day in Microsoft Outlook — and comes amid important security updates from other software makers, including Adobe, Apple, Google and Mozilla.

Image: Shutterstock.com. Diorama by @zepp1969.

November’s patch release from Redmond tackles 1,093 vulnerabilities, fixing 63 “critical” flaws that can be exploited to seize control over a Windows system with little or no help from users. Nearly half of those critical bugs are remote code execution vulnerabilities. Overall, 535 of the flaws addressed this month allow an attacker to elevate their privileges on a vulnerable device.

This latest round of updates means Microsoft has shipped more than 5,100 security patches so far this year, a staggering number considering that a year ago Microsoft was averaging around 100 patches per month. Microsoft credits advances in artificial intelligence for helping to find (and fix) such a large volume of security bugs in recent months. Last month, Microsoft pushed more than 1,400 fixes.

According to Redmond, just one of the flaws is already seeing active exploitation: CVE-2026-124194, a spoofing bug in Outlook. The advisory for CVE-2026-124194 says it affects Microsoft 365 Apps for Enterprise and editions of Microsoft Office.

Several other major software makers pushed security updates this week to fix critical weaknesses in their products. Apple on Monday released out-of-cycle security updates to fix at least nine vulnerabilities in iOS/iPadOS, macOS, watchOS and visionOS. Among the weaknesses reportedly fixed in Apple’s release include an iMessage flaw that could let a remote attacker join a group chat and receive its messages; the updates also fix a separate iMessage flaw that could let someone added to a group access messages sent before they joined.

Lee Neely, an instructor at the SANS Institute, noted that Apple backported the fix to iOS 18.

“I take notice when Apple releases fixes to more than two versions,” Neely wrote in a SANS newsletter today. “You really need to get to version 27.2.1 of their OSes and push folks who are holding onto devices which cannot run the current version. The best plan often involves setting an upgrade date for old hardware. Run through all your Apple devices and make sure everything is up to date, don’t assume the fixes went out automatically.”

Google this week released Chrome version 157.0.8103.50, which resolves nine security holes, including CVE-2026-125350, the sixth zero-day vulnerability this year for Chrome (this bug also being patched in Chromium-based browsers like Edge). The latest Chrome version follows its November 3 release, which fixed 213 vulnerabilities. A new version of Mozilla Firefox out today (Firefox 160) resolves almost 100 CVEs, about half of them high severity.

Adobe today released patches for at least 110 vulnerabilities across a range of products, with nearly half of the flaws rated high severity. Adobe said there is no evidence any of the security holes it addressed this month are being actively exploited.

Chris Goettl, vice president of product management at Ivanti, said it’s easy for security teams at large enterprises to feel overwhelmed by the volume and criticality of patches from Microsoft lately. But he notes that while some 95,000 vulnerabilities have been assigned CVEs this year across all vendors, only 217 have so far this year been added to the Known Exploited Vulnerabilities (KEV) list maintained by the U.S. Cybersecurity and Infrastructure Security Agency (CISA).

“Roughly four in ten of this year’s exploited vulnerabilities sit in operating systems, endpoint software, browsers, and server platforms, which a continuous patch strategy can close quickly,” Goettl said.”The rest calls for exposure management alongside patching, especially the 44% that sit in edge devices and IT management or remote access tools, the internet-facing systems that hold privileged access. Teams that can quickly identify those assets, confirm exposure and act on confirmed exploitation can focus on the small slice of vulnerabilities attackers are actively using, while the rest stays on the normal patch cadence.”

Please don’t neglect to back up any important data if you haven’t recently; generally speaking, this is a task best accomplished before applying operating system updates. As ever, if you experience any issues applying this month’s bundle of love from Microsoft, please drop a comment about it in the space below.

Article URL: https://msrc.microsoft.com/update-guide/vulnerability/CVE-2026-116928

Comments URL: https://news.ycombinator.com/item?id=50446357

Points: 2

# Comments: 0

I am an Android developer am new here any welecome

Comments URL: https://news.ycombinator.com/item?id=50446341

Points: 1

# Comments: 0

Article URL: https://shellport.app/

Comments URL: https://news.ycombinator.com/item?id=50446320

Points: 1

# Comments: 0

Article URL: https://pietromiotti.com/latest/maya-site/

Comments URL: https://news.ycombinator.com/item?id=50446316

Points: 1

# Comments: 0

Article URL: https://brian.biz/blog/ai-agents-verify/

Comments URL: https://news.ycombinator.com/item?id=50446314

Points: 1

# Comments: 0

Article URL: https://github.com/TinkeringDao/satori

Comments URL: https://news.ycombinator.com/item?id=50446310

Points: 1

# Comments: 0

Title.

Comments URL: https://news.ycombinator.com/item?id=50446309

Points: 1

# Comments: 1

Article URL: https://www.wsj.com/us-news/education/the-parents-suing-schools-because-their-kids-got-bad-grades-c6c12415

Comments URL: https://news.ycombinator.com/item?id=50446307

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/playlist?list=PLUfqNFdN-UYk

Comments URL: https://news.ycombinator.com/item?id=50446293

Points: 2

# Comments: 0

Article URL: https://migflug.com/afterburner/fighter-radomes-explained-radar-transparent-nose-cone-boresight-rain-erosion/

Comments URL: https://news.ycombinator.com/item?id=50446266

Points: 2

# Comments: 0

Article URL: https://teorth.github.io/tao-web/six-math-essentials.html

Comments URL: https://news.ycombinator.com/item?id=50446246

Points: 9

# Comments: 1

Article URL: https://twitter.com/aaronp613/status/2120229190477992162

Comments URL: https://news.ycombinator.com/item?id=50446240

Points: 5

# Comments: 0