Grok told Trump to invade Venezuela, ChatGPT had a security flaw, and Suno now speaks

Quick question before we dive in: when was the last time an AI chatbot allegedly helped shape a geopolitical decision at the highest level of government? Because that may have just happened.

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

This one is wild. According to a TechCrunch report, President Trump reportedly consulted Elon Musk's Grok AI chatbot before making decisions about Venezuela - and Grok apparently encouraged him to capture President Nicolás Maduro.

The incident raises an uncomfortable question that nobody in Washington seems to be asking loudly enough: should a commercial AI chatbot have any role - even an informal one - in shaping military or foreign policy decisions?

This connects to the broader pattern we've been tracking around AI and military decision-making. If you've been following our coverage of AI agents in high-stakes environments, the implications here are significant - we're no longer talking about hypothetical risks.

Grok is built by xAI, Musk's AI company, and is designed to be less filtered than competitors. That design choice looks very different when the person asking questions is the President of the United States.

Read the full story →

An AI chatbot with minimal guardrails being consulted on foreign policy is not a fringe scenario anymore - it's apparently already happening, and there's no formal framework in place to prevent it.

The U.S. cattle herd has fallen to 86.2 million head. 

That's the lowest level since 1951. And the problem isn't going away, plus ranchers face rising costs and a tightening labor market. 

Yet America still needs to produce beef. 

The equation is simple: Get more from every animal. 

DIT AgTech helps ranchers produce more with less through AI, nutrition automation and real-time data. Its infrastructure boosts herd performance while reducing labor and costs. 

Proven at scale in Australia: 

What drives adoption? Ranchers get the technology free with a three-year nutrition plan. 

Removing the cost barrier allows DIT AgTech to scale adoption faster, generating more data and recurring revenue 

America needs more output from fewer cattle. DIT AgTech has built the infrastructure to make that possible.

Invest in DIT AgTech Today

^{𝘐𝘯 𝘮𝘢𝘬𝘪𝘯𝘨 𝘢𝘯 𝘪𝘯𝘷𝘦𝘴𝘵𝘮𝘦𝘯𝘵 𝘥𝘦𝘤𝘪𝘴𝘪𝘰𝘯, 𝘪𝘯𝘷𝘦𝘴𝘵𝘰𝘳𝘴 𝘮𝘶𝘴𝘵 𝘳𝘦𝘭𝘺 𝘰𝘯 𝘵𝘩𝘦𝘪𝘳 𝘰𝘸𝘯 𝘦𝘹𝘢𝘮𝘪𝘯𝘢𝘵𝘪𝘰𝘯 𝘰𝘧 𝘵𝘩𝘦 𝘪𝘴𝘴𝘶𝘦𝘳 𝘢𝘯𝘥 𝘵𝘩𝘦 𝘵𝘦𝘳𝘮𝘴 𝘰𝘧 𝘵𝘩𝘦 𝘰𝘧𝘧𝘦𝘳𝘪𝘯𝘨, 𝘪𝘯𝘤𝘭𝘶𝘥𝘪𝘯𝘨 𝘵𝘩𝘦 𝘮𝘦𝘳𝘪𝘵𝘴 𝘢𝘯𝘥 𝘳𝘪𝘴𝘬𝘴 𝘪𝘯𝘷𝘰𝘭𝘷𝘦𝘥. 𝘋𝘐𝘛 𝘈𝘨𝘛𝘦𝘤𝘩 𝘩𝘢𝘴 𝘧𝘪𝘭𝘦𝘥 𝘢 𝘍𝘰𝘳𝘮 𝘊 𝘸𝘪𝘵𝘩 𝘵𝘩𝘦 𝘚𝘦𝘤𝘶𝘳𝘪𝘵𝘪𝘦𝘴 𝘢𝘯𝘥 𝘌𝘹𝘤𝘩𝘢𝘯𝘨𝘦 𝘊𝘰𝘮𝘮𝘪𝘴𝘴𝘪𝘰𝘯 𝘪𝘯 𝘤𝘰𝘯𝘯𝘦𝘤𝘵𝘪𝘰𝘯 𝘸𝘪𝘵𝘩 𝘪𝘵𝘴 𝘰𝘧𝘧𝘦𝘳𝘪𝘯𝘨, 𝘢 𝘤𝘰𝘱𝘺 𝘰𝘧 𝘸𝘩𝘪𝘤𝘩 𝘮𝘢𝘺 𝘣𝘦 𝘰𝘣𝘵𝘢𝘪𝘯𝘦𝘥 𝘩𝘦𝘳𝘦: https://bit.ly/4bzuWCi​}

While most headlines focus on what AI can hack, a newly patched vulnerability in ChatGPT's Mac app flips that story: AI software itself is an attractive target for attackers.

The flaw, reported by Wired, could have allowed hackers to grab sensitive data from users of the ChatGPT desktop app on Mac. The vulnerability has since been patched by OpenAI, but its existence points to a growing attack surface that doesn't get nearly enough attention.

As AI apps become deeply embedded in daily workflows - handling documents, code, emails, and conversations - a single app-level vulnerability can expose far more than a typical software bug. The blast radius is bigger because the data flowing through these apps is more sensitive.

If you're using AI desktop apps, especially ones with access to local files or memory features, it's worth checking that you're running the latest version. This won't be the last vulnerability of this kind.

Read the full story →

Update your AI apps like you update your browser - these aren't just productivity tools anymore, they're privileged access points to your most sensitive data.

Suno built its reputation generating AI music, but today it's doing something new: generating spoken voices. The feature - now in public beta on Suno's web and mobile platforms - lets you create voiceovers from scripts or prompted descriptions, with background music generated simultaneously.

The pitch is clear: instead of going to one tool for voice and another for music, Suno handles both at once. That's a genuinely useful workflow for content creators, podcast intro producers, or anyone building short video content.

It puts Suno in more direct competition with voice-first platforms like ElevenLabs - which just doubled its valuation to $22 billion this week - though Suno's angle is the music-plus-speech combination rather than pure voice quality. The AI audio space is getting crowded fast.

If you're a creator who has been experimenting with AI audio tools, this is worth a look while it's in free beta. These windows tend to close once pricing kicks in.

Read the full story →

Suno is evolving from a music generator into a full audio production platform - and the combined voice-plus-music feature is a real differentiator for video and podcast creators.

There's a sharp gap between the promise of healthcare AI and the daily reality for the workers using it. Wired's investigation into Palantir's hospital scheduling software is one of the clearest examples yet.

Healthcare workers at a hospital giant and radiology network say Palantir's scheduling system is causing errors, increasing burnout, and creating frustration across nursing and clinical staff. Instead of streamlining workloads, the software is reportedly generating new administrative burdens that humans then have to clean up.

This fits a pattern we've seen emerging across healthcare AI deployments - the technology optimizes for metrics that look good on paper while missing the messy human realities of clinical work. If you read our earlier coverage of the $942M healthcare AI problem, this story is more of the same pressure from a different angle.

The healthcare AI space attracted enormous investment on the promise of reducing burnout. The irony of it adding to burnout instead is not lost on the workers living it.

Read the full story →

AI that optimizes for scheduling efficiency without input from the humans doing the work doesn't just fail - it actively makes things worse, and someone still has to fix it manually.

A new AI tool can analyze a brain scan and reconstruct - with remarkable accuracy - what a person was looking at when the scan was taken. MIT Technology Review reports the system works in both directions: show it a brain scan and it guesses the image, or show it an image and it predicts the brain activity pattern.

The research sits at the intersection of neuroscience and machine learning, and the potential applications run from medical (helping locked-in patients communicate) to deeply concerning (unauthorized surveillance of mental imagery, eventually).

We're not at real-time mind-reading yet - this requires fMRI scanners that are bulky and expensive. But the direction of travel is clear, and the accuracy improvements are happening faster than the ethical frameworks around it.

Read the full story →

The gap between 'impressive research demo' and 'real-world application' is narrowing faster than regulators can track - this technology deserves a policy conversation now, not after it's widely deployed.

By the way - if anything in today's newsletter has you thinking about updating your online presence or building a quick project page, check out 60sec.site. It's an AI-powered website builder that gets you from zero to live in about a minute. No coding needed. And for daily AI news coverage, dailyinference.com is where we publish everything.

The answer is Suno! The AI music platform launched spoken voice generation in public beta this week, letting users create voiceovers and background music simultaneously across web and mobile platforms.

The Grok-and-Trump story raises something I'm genuinely curious about: do you think AI chatbots should have any role in helping leaders make political or military decisions, even informally? Hit reply and tell me where you land on this - I read every response and this one feels like it could go a lot of directions.

That's it for today's edition. A lot happening across safety, tools, and some genuinely unsettling use cases - the week is ending on a complicated note. See you Monday with more. Have a good weekend.

This newsletter grows when readers pass it on. Share your link and you get the AI Tools Starter Kit at one referral, and the AI Insider Briefing at five.