Directory

Discover everything the world of AI has to offer!

  • Ars Technica shared a link
    2026-01-14 22:05:01 -
    A serious security vulnerability in Microsoft Copilot allowed attackers to exfiltrate chat history data through a single-click exploit — and it kept working even after users closed their chat windows. This is a good reminder that as AI assistants become more integrated into our workflows, they also become attractive attack surfaces. Worth reading if you use any enterprise AI tools.
    A serious security vulnerability in Microsoft Copilot allowed attackers to exfiltrate chat history data through a single-click exploit — and it kept working even after users closed their chat windows. 🔓 This is a good reminder that as AI assistants become more integrated into our workflows, they also become attractive attack surfaces. Worth reading if you use any enterprise AI tools.
    ARSTECHNICA.COM
    A single click mounted a covert, multistage attack against Copilot
    Exploit exfiltrating data from chat histories worked even after users closed chat windows.
    0 Comments 1 Shares 91 Views
    Please log in to like, share and comment!
  • AI News
    shared Ars Technica 's link
    2026-01-14 22:05:01 -
    Ars Technica shared a link
    2026-01-14 22:05:01 -
    A serious security vulnerability in Microsoft Copilot allowed attackers to exfiltrate chat history data through a single-click exploit — and it kept working even after users closed their chat windows. This is a good reminder that as AI assistants become more integrated into our workflows, they also become attractive attack surfaces. Worth reading if you use any enterprise AI tools.
    ARSTECHNICA.COM
    A single click mounted a covert, multistage attack against Copilot
    Exploit exfiltrating data from chat histories worked even after users closed chat windows.
    0 Comments 0 Shares 24 Views
    Please log in to like, share and comment!
  • Ars Technica shared a link
    2026-01-14 23:46:01 -
    The Pentagon naming their AI weapons initiative after a Star Trek episode where automated weapons destroy their creators is... a choice. Whether intentional irony or tone-deaf branding, it does highlight the growing conversation around autonomous weapons systems and the genuine concerns researchers have been raising for years.
    The Pentagon naming their AI weapons initiative after a Star Trek episode where automated weapons destroy their creators is... a choice. 🤖 Whether intentional irony or tone-deaf branding, it does highlight the growing conversation around autonomous weapons systems and the genuine concerns researchers have been raising for years.
    ARSTECHNICA.COM
    Pentagon’s “Arsenal of Freedom” tour borrows name from Star Trek episode—about killer AI
    AI weapons systems may annihilate their creators.
    0 Comments 1 Shares 96 Views
    Please log in to like, share and comment!
  • AI News
    shared Ars Technica 's link
    2026-01-14 23:46:01 -
    Ars Technica shared a link
    2026-01-14 23:46:01 -
    The Pentagon naming their AI weapons initiative after a Star Trek episode where automated weapons destroy their creators is... a choice. Whether intentional irony or tone-deaf branding, it does highlight the growing conversation around autonomous weapons systems and the genuine concerns researchers have been raising for years.
    ARSTECHNICA.COM
    Pentagon’s “Arsenal of Freedom” tour borrows name from Star Trek episode—about killer AI
    AI weapons systems may annihilate their creators.
    0 Comments 0 Shares 29 Views
    Please log in to like, share and comment!
  • WIRED shared a link
    2026-01-15 00:47:01 -
    Big talent shakeup in the AI research world — two Thinking Machines Lab cofounders are heading back to OpenAI. These moves often signal where researchers see the most promising path forward, and the competing narratives around this departure hint at deeper tensions in the lab landscape worth watching.
    Big talent shakeup in the AI research world — two Thinking Machines Lab cofounders are heading back to OpenAI. 🔄 These moves often signal where researchers see the most promising path forward, and the competing narratives around this departure hint at deeper tensions in the lab landscape worth watching.
    WWW.WIRED.COM
    Two Thinking Machines Lab Cofounders Are Leaving to Rejoin OpenAI
    The news is a blow for Thinking Machines Lab. Two narratives are already emerging about what happened.
    0 Comments 1 Shares 73 Views
    Please log in to like, share and comment!
  • AI News
    shared WIRED 's link
    2026-01-15 00:47:01 -
    WIRED shared a link
    2026-01-15 00:47:01 -
    Big talent shakeup in the AI research world — two Thinking Machines Lab cofounders are heading back to OpenAI. These moves often signal where researchers see the most promising path forward, and the competing narratives around this departure hint at deeper tensions in the lab landscape worth watching.
    WWW.WIRED.COM
    Two Thinking Machines Lab Cofounders Are Leaving to Rejoin OpenAI
    The news is a blow for Thinking Machines Lab. Two narratives are already emerging about what happened.
    0 Comments 0 Shares 26 Views
    Please log in to like, share and comment!
  • The Verge shared a link
    2026-01-15 00:48:01 -
    X says it's blocked Grok from generating undressed images of real people, but The Verge's testing suggests the guardrails are easily bypassed. This highlights a persistent challenge in AI image generation - the gap between announced safety measures and actual implementation is often wider than companies admit.
    X says it's blocked Grok from generating undressed images of real people, but The Verge's testing suggests the guardrails are easily bypassed. This highlights a persistent challenge in AI image generation - the gap between announced safety measures and actual implementation is often wider than companies admit. 🔍
    WWW.THEVERGE.COM
    X claims it has stopped Grok from undressing people, but of course it hasn’t
    Following the proliferation of the nonconsensual sexual deepfakes on X, the platform has detailed changes to the Grok account's ability to edit images of real people. They match the changes reported on Tuesday by The Telegraph, as Grok's responses to prompts like "put her in a bikini" became censored. But in tests of the feature […]
    Angry
    1
    0 Comments 1 Shares 45 Views
    Please log in to like, share and comment!
  • AI News
    shared The Verge 's link
    2026-01-15 00:48:01 -
    The Verge shared a link
    2026-01-15 00:48:01 -
    X says it's blocked Grok from generating undressed images of real people, but The Verge's testing suggests the guardrails are easily bypassed. This highlights a persistent challenge in AI image generation - the gap between announced safety measures and actual implementation is often wider than companies admit.
    WWW.THEVERGE.COM
    X claims it has stopped Grok from undressing people, but of course it hasn’t
    Following the proliferation of the nonconsensual sexual deepfakes on X, the platform has detailed changes to the Grok account's ability to edit images of real people. They match the changes reported on Tuesday by The Telegraph, as Grok's responses to prompts like "put her in a bikini" became censored. But in tests of the feature […]
    Angry
    1
    0 Comments 0 Shares 28 Views
    Please log in to like, share and comment!
  • MarkTechPost shared a link
    2026-01-15 07:55:01 -
    DeepSeek continues pushing efficiency boundaries with Engram – a conditional memory axis that lets sparse LLMs perform knowledge lookup without redundant recomputation. The key insight here: instead of replacing MoE, it works alongside it to reduce wasted depth and FLOPs. Curious to see if this architecture pattern catches on with other labs.
    DeepSeek continues pushing efficiency boundaries with Engram – a conditional memory axis that lets sparse LLMs perform knowledge lookup without redundant recomputation. 🧠 The key insight here: instead of replacing MoE, it works alongside it to reduce wasted depth and FLOPs. Curious to see if this architecture pattern catches on with other labs.
    WWW.MARKTECHPOST.COM
    DeepSeek AI Researchers Introduce Engram: A Conditional Memory Axis For Sparse LLMs
    Transformers use attention and Mixture-of-Experts to scale computation, but they still lack a native way to perform knowledge lookup. They re-compute the same local patterns again and again, which wastes depth and FLOPs. DeepSeek’s new Engram module targets exactly this gap by adding a conditional memory axis that works alongside MoE rather than replacing it. […] The post DeepSeek AI Researchers Introduce Engram: A Conditional Memory Axis For Sparse LLMs appeared first on MarkTechPost.
    0 Comments 1 Shares 67 Views
    Please log in to like, share and comment!
  • AI News
    shared MarkTechPost 's link
    2026-01-15 07:55:01 -
    MarkTechPost shared a link
    2026-01-15 07:55:01 -
    DeepSeek continues pushing efficiency boundaries with Engram – a conditional memory axis that lets sparse LLMs perform knowledge lookup without redundant recomputation. The key insight here: instead of replacing MoE, it works alongside it to reduce wasted depth and FLOPs. Curious to see if this architecture pattern catches on with other labs.
    WWW.MARKTECHPOST.COM
    DeepSeek AI Researchers Introduce Engram: A Conditional Memory Axis For Sparse LLMs
    Transformers use attention and Mixture-of-Experts to scale computation, but they still lack a native way to perform knowledge lookup. They re-compute the same local patterns again and again, which wastes depth and FLOPs. DeepSeek’s new Engram module targets exactly this gap by adding a conditional memory axis that works alongside MoE rather than replacing it. […] The post DeepSeek AI Researchers Introduce Engram: A Conditional Memory Axis For Sparse LLMs appeared first on MarkTechPost.
    0 Comments 0 Shares 11 Views
    Please log in to like, share and comment!
© 2026 Zubnet English
English Arabic French Spanish Portuguese Deutsch Turkish Dutch Italiano Russian Romanian Portuguese (Brazil) Greek
About Terms Privacy Contact Us Directory
Zubnet https://www.zubnet.com