Today's Hook
Researchers discovered that hackers can hide inaudible sounds inside a podcast or YouTube video, sounds below the range of human hearing, that silently hijack your phone's AI assistant. Once the attack runs, it can access your photos, bank accounts, and anything else connected to your voice AI.
You do not have to do anything wrong. The audio just plays in the background. The attack takes about 30 minutes to build, and it works regardless of what you were saying or doing when it hit you.
Your move, Siri.
Today in AI
What Happened While You Were Sleeping
A free GitHub tool bypassed key safety guardrails on Meta and Google's AI models in under 10 minutes.
ClickUp fired 22% of its staff and replaced them with 3,000 AI agents.
Grok's next model finished training. Elon Musk says it goes public in 2-3 weeks.
California State University doubled down on a $13M/year OpenAI deal, even as its own faculty and students push back.
From Our Partners
If you're a consultant, you probably know the feeling:
A client meeting is coming up.
You know the slide exists.
You just cannot find it.
So you dig through old decks, random folders, and SharePoint - or rebuild it from scratch.
SlideHub helps consultants and sales reps find the right approved slide faster, so less time goes into the hunt and more goes into the deck.
😸 Meta Spent Millions on AI Safety. A Free GitHub Tool Bypassed Much of It in 10 Minutes.
The AI industry's uncomfortable open secret just got a lot harder to ignore.
Meta and Google have spent hundreds of millions building safety guardrails into their AI models, the filters that stop those models from explaining how to make weapons or generate malware. Last week, a Financial Times investigation found that a free tool called Heretic, available on GitHub, bypassed key safeguards in under 10 minutes. On a regular laptop.
The modified model then answered questions about biological weapons it had previously refused to discuss.
Here is what happened
The FT used Heretic to strip safety filters from Meta's Llama 3.3 in under 10 minutes, no special hardware needed.
A separate test on Google's Gemma 3 model produced similarly alarming outputs, including instructions the original model would have refused.
Heretic's creator told the FT the tool has already been used to build over 3,500 "decensored" model versions, downloaded 13 million times.
He also bypassed Google's newer Gemma 4 model within 90 minutes of its public release.
Here is the key thing to understand: this technique (called "abliteration") only works on open-source models, meaning models where anyone can download and modify the underlying code. Proprietary models like Claude or ChatGPT are harder targets because outsiders cannot access those core files directly.
The pattern is broader than one tool. A Nature Communications study found that reasoning-capable AI models could autonomously talk other AI models into producing harmful outputs, with a 97% success rate across major commercial models. An ICLR 2026 paper described a more surgical approach with up to 99% bypass rates on some models.
The uncomfortable lesson is not that one GitHub tool is uniquely dangerous. It is that open-weight AI changes the safety equation completely. Safety stops being a locked door and starts being more like a sticker that determined users can peel off.
Ai Skill Of The Day
Most AI debates miss the point entirely. The question is not "Copilot vs. Gemini vs. Claude." It is "which one lives where you already work?"
Here is the framework that actually matters, broken down cleanly:
Use Copilot if your team runs on Microsoft 365. It is native inside Word, Excel, Outlook, Teams, and GitHub. Many enterprise companies block ChatGPT but allow Copilot, making it the most-adopted AI tool in corporate settings whether anyone admits it or not.
Use Gemini if your work lives in Google Workspace. Gmail, Docs, Sheets, Drive: Gemini is built into all of it. Best for summarizing email threads, drafting slides, and the async collaboration that eats half your workday.
Use Claude when the task requires real thinking across large amounts of material: legal review, research synthesis, long-document analysis, or anything where you need careful reasoning rather than fast execution.
Routing the right task to the right model is itself a skill. Most people are not doing it. The best AI is not the most popular one. It is the one that integrates into how your team already works.
What's Trending
ClickUp cut 22% of its workforce (about 290 people) and replaced them with 3,000 AI agents, calling it a "100x org"; surviving employees are being offered salary bands up to $1M if they create outsized impact using AI.
Grok's next model (V9-Medium, a 1.5 trillion parameter model) finished training with strong early results; fine-tuning is underway with public release about 2-3 weeks out.
California State University renewed its $13M/year OpenAI deal (a 3-year, $39M+ commitment) to become the first AI-powered university system in the US, even as a majority of its own students and faculty said they are skeptical of AI's educational value.
Cybersecurity job postings jumped 11% year-over-year in Q1 2026 as AI-generated code flooded the market with new vulnerabilities, making it one of the few job categories actively growing because of AI, not despite it.
LA's sidewalk delivery robots expanded to 40 neighborhoods (up from just 2 in 2023) as Serve Robotics grew its fleet elevenfold; local restaurants describe the bots as a daily fixture that "everyone films."
Prompt Station
Paste this into Claude, Copilot, or Gemini to figure out which AI tool belongs in your workflow:
I want to figure out which AI tool fits my work best. Here is how I work: - My main tools are: [list your apps, e.g. Gmail, Outlook, Notion, Slack] - My most time-consuming tasks are: [e.g. writing reports, summarizing meetings, answering emails] - My team uses: [Google Workspace / Microsoft 365 / other] Based on this, which AI assistant should I use for each task type, and why? Give me a simple one-paragraph recommendation I can act on today.
Holmris saves 80% of their deck creation time with SlideHub. The AI-powered slide library used by KPMG, Netcompany, and 500+ firms, built inside PowerPoint.