<h2>Must know</h2>
<ul>
<li><strong>Microsoft responds to AI data center backlash</strong>: Facing significant public opposition across 24 states, Microsoft is changing its approach to AI infrastructure development. The company announced a “community-first” plan, committing to pay full electricity grid costs, minimize water usage, and create local jobs, as reported by multiple sources. This initiative comes after Microsoft withdrew from a Wisconsin data center project and faces ongoing resistance in Michigan and Ohio, highlighting the growing environmental and social impact concerns of AI’s energy demands. <a href="https://app.daily.dev/posts/jqtBl3Jnb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical vulnerabilities found in AI/ML Python libraries</strong>: Security researchers discovered remote code execution (RCE) vulnerabilities in widely used AI/ML Python libraries from Nvidia, Salesforce, and Apple. These flaws, stemming from unsafe use of Hydra’s <code>instantiate()</code> function, allow attackers to embed malicious code in model metadata that executes when loaded. Fixes have been issued, but the issue highlights a significant supply chain attack vector in AI/ML workflows, especially given the millions of downloads from platforms like Hugging Face. <a href="https://app.daily.dev/feed-by-ids?id=4tbUFTlCZ&amp;id=c7PnFTJdg&amp;id=kZzcz2YmW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pentagon integrates Grok AI despite past controversies</strong>: US Defense Secretary Pete Hegseth announced the Pentagon’s intent to integrate Elon Musk’s Grok AI into both classified and unclassified networks. This move is part of a broader AI acceleration strategy, despite Grok AI’s past controversies regarding inappropriate image generation. The Pentagon is also facing criticism for using outdated assessments for quantum technology planning, creating uncertainty for vendors and potentially hindering military operational readiness, according to a Defense Department Inspector General report. <a href="https://app.daily.dev/feed-by-ids?id=KoB4l9TcK&amp;id=jRGkCZr4W" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Quantum tech security and commercial developments accelerate</strong>: Penn State researchers warn that cloud-based quantum computers have significant security vulnerabilities, including intellectual property theft and undetectable result degradation, urging immediate security integration. Meanwhile, IBM released its 120-qubit Nighthawk processor, Rigetti delayed its 108-qubit Cepheus-1-108Q, and UC Berkeley demonstrated a three-qubit silicon photonic chip, while New York State plans to expand its quantum commercialization hubs. <a href="https://app.daily.dev/feed-by-ids?id=JmIuGeLwy&amp;id=MkzC4PEYB&amp;id=QPb7M1IyK&amp;id=dtnIdUSBt&amp;id=f8i188R5R&amp;id=hhNibwkwa&amp;id=yjM5va3Id&amp;id=Lx6jVDI9P&amp;id=RONvWGnCv&amp;id=bD5nKoPwD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta’s AI leadership shakeup and book spam</strong>: Yann LeCun, Meta’s former chief AI scientist, departed after Alexander Wang was appointed head of superintelligence, with LeCun publicly criticizing Wang’s experience and alleging manipulation of Llama 4 test results. Separately, HiTeX Press has published over 800 AI-generated technical books, many with fabricated content, highlighting a growing problem of industrial-scale AI book spam. <a href="https://app.daily.dev/feed-by-ids?id=ILAkYeoup&amp;id=PmK6l9z8J" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Brazil halts Meta’s WhatsApp AI chatbot ban</strong>: Brazil’s antitrust regulator CADE ordered Meta to suspend its new WhatsApp Business API policy that prohibits third-party AI chatbots, investigating whether the move gives Meta AI an unfair competitive advantage. <a href="https://app.daily.dev/posts/HP1IfGhyt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Advanced VoidLink Linux malware targets cloud</strong>: Check Point Research identified VoidLink, an advanced Linux malware written in Zig, designed to target cloud and container environments like Docker and Kubernetes, featuring 37+ modular plugins for credential harvesting and evasion. <a href="https://app.daily.dev/feed-by-ids?id=uJaIfhgHH&amp;id=QWUKVWdsK&amp;id=x9wl564Rc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI reshapes open source business models</strong>: AI systems are disrupting documentation-centric open source business models by extracting and caching documentation, leading to significant revenue declines for projects like Tailwind CSS, which saw an 80% drop. <a href="https://app.daily.dev/feed-by-ids?id=Ml47Y2Hd6&amp;id=l6Y9rCI4U" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js releases security fixes</strong>: Node.js released versions 20.20.0 (LTS) and 25.3.0 (Current) with security fixes addressing multiple CVEs, including a denial-of-service vulnerability from stack space exhaustion impacting users of <code>async_hooks</code> and React Server Components. <a href="https://app.daily.dev/feed-by-ids?id=DwsoMgMKH&amp;id=ImHNcBPhJ&amp;id=ZoKLGVJ9V&amp;id=uw83EGG5r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New York advances robotaxi legislation</strong>: New York Governor Kathy Hochul proposed legislation to legalize commercial robotaxi operations statewide, explicitly excluding New York City, and removing the requirement for a human safety driver. <a href="https://app.daily.dev/posts/HMVvWNGfO" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Slackbot is now an ‘AI agent’ for enterprise customers</strong>: Salesforce has ‘re-released’ Slackbot as an ‘AI agent’ for its Business+ and Enterprise+ customers. Apparently, this means it can now ‘answer questions, organize tasks, create content, and schedule meetings’—you know, all the groundbreaking stuff bots have been doing for years. The real innovation here seems to be limiting access to higher-tier customers and calling it ‘AI’ to justify the price tag. Don’t worry, no installation required, just your eternal subscription. <a href="https://app.daily.dev/posts/VpXLIsjSp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents will fail, but don’t worry, Gartner has a plan</strong>: Gartner, ever the oracle of obviousness, predicts 40% of AI agent projects will fail by 2027, mostly because people aren’t putting in ‘guardrails.’ Who knew that giving an AI agent free rein to sell a $76,000 car for $1 (looking at you, Chevrolet chatbot) might be a bad idea? Their solution? Treat AI agents like employees, with ‘governance, technical controls, and human oversight.’ Revolutionary! It’s almost like they’re suggesting we apply basic engineering principles to our ‘super-intelligent’ systems. <a href="https://app.daily.dev/posts/QnCK3ryB4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shadow AI is the new shadow API, apparently</strong>: Nordic APIs is here to tell us that ‘Shadow AI’ is the new ‘Shadow API,’ creating security risks because employees are using AI tools outside IT oversight. Shocking, I know. Apparently, these AI systems ‘ingest new data and operate with weak or nonexistent identity controls,’ leading to ‘identity collapse.’ It’s almost like giving powerful tools to people without proper governance is a bad idea, regardless of whether it’s an API or an ‘AI.’ Who could have predicted? <a href="https://app.daily.dev/posts/yF1D00gr0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
