Ivy: Without deliberate correction, AI will simply give old discrimination new speed and authority.
Marcus: I'm Marcus.
Ivy: And I'm Ivy.
Marcus: And this is AI Unfiltered for Tuesday, August 25th, 2026.
Ivy: On the show today: how AI is changing our world, and not always for the better.
Marcus: We'll start with that chilling warning from the United Nations about AI supercharging bias.
Ivy: Then, a new tool from JetBrains that keeps your code private by running an AI assistant right on your own machine.
Marcus: And we'll dig into a new report that says your most productive employees might also be your biggest AI security risk.
Ivy: So let's unpack that warning from the UN's Deputy Secretary-General, Amina J. Mohammed.
Marcus: Right. She's saying that the default trajectory for AI isn't some new utopia, it's just automating our oldest, worst habits.
Ivy: And she was precise. She said 'without deliberate correction.' That's not a passive warning. It means the default setting is biased, and we have to actively fight it.
Marcus: But aren't the big labs already working on this? Fairness and alignment are huge buzzwords.
Ivy: Are they working on it, or are they marketing that they're working on it? The UN's point is that the speed and scale make this a whole new kind of threat.
Marcus: Explain that. What's the difference?
Ivy: A single biased loan officer in the 1980s could ruin a few hundred lives. A single biased algorithm in 2026 can ruin millions in a second, and we'd call it 'efficiency'.
Marcus: Oof. Okay, that's a gut punch. So what's the 'deliberate correction' they're asking for?
Ivy: Actual regulation. Global standards. Things that have teeth, beyond just hoping tech companies do the right thing when it's not profitable.
Marcus: So for builders, the takeaway is 'garbage in, garbage out' is a social problem now, not just a technical one. You have to actively de-bias your data.
Ivy: And for everyone else, it means you should probably start questioning the computer when it says 'no'.
Marcus: Alright, let's shift gears to something a bit more optimistic—a tool that might solve a piece of this puzzle. Tell us about Junie Local.
Ivy: It's JetBrains' new AI coding assistant, and the whole point is that it runs entirely on your local machine—a Mac, for now. Your code never leaves your computer.
Marcus: Which is a game-changer! So many companies have banned tools like GitHub Copilot over fears of sending proprietary code to a third party. This sidesteps that completely.
Ivy: It solves the privacy problem, yes. But it introduces a performance question. The models running on your laptop aren't going to be as powerful as the giant ones in a datacenter.
Marcus: Right, and JetBrains is upfront about that. They're not claiming it's a GPT-5 killer. It's built for those high-context, low-latency tasks—like smart code completion, refactoring, or generating tests for the file you're working on right now.
Ivy: So it's less of a creative partner and more of a very, very smart assistant with a perfect short-term memory.
Marcus: Exactly! Which, for day-to-day coding, is arguably more useful than an AI that can write a sonnet about JavaScript but can't find a bug in my code.
Ivy: So for developers, this means you have an option your paranoid CISO might actually approve. Just don't expect it to architect your entire application for you.
Marcus: It's the ultimate trade-off: peak intelligence versus peak privacy. For enterprise, privacy usually wins.
Ivy: Speaking of paranoid CISOs, they have a new reason to worry. A new Akamai report suggests your best employees might be your worst enemies on security.
Marcus: Okay, as an enthusiastic early adopter, I feel seen. This is about 'super-adopters,' right?
Ivy: Exactly. The report finds that just 5% of employees—these 'super-adopters'—account for a huge chunk of a company's AI use. And they're creating massive security holes.
Marcus: But how? We're the ones driving productivity! We're trying to help!
Ivy: By using your personal, free AI accounts for work. You're pasting proprietary source code, sensitive customer data, and internal marketing plans into public-facing tools the company has zero control over.
Marcus: Okay, when you put it like that... it sounds bad. But it's just the new 'Shadow IT,' right? People trying to get their jobs done with better tools.
Ivy: The report calls it 'Shadow AI,' and it's far more dangerous. With Shadow IT, you might sync a file to a personal Dropbox. With Shadow AI, you're actively feeding your company's crown jewels to a third-party model as training data.
Marcus: So you're not just storing it there, you're... teaching the world your secrets. That's a real 'oh sh—' moment.
Ivy: The takeaway isn't to punish these users. It's for security teams to get ahead of the problem. Provide safe, sanctioned tools—like that Junie Local we just talked about—and explain why pasting the Q4 strategy into a free chatbot is a terrible idea.
Marcus: Embrace and empower, don't block and blame. Find your super-adopters and make them champions for secure AI, not your biggest risk.
Ivy: Exactly. Give them a sandbox before they set the building on fire.
Marcus: Alright, so let's tie this all together. The UN is warning that without guardrails, AI is set to accelerate discrimination.
Ivy: Meanwhile, JetBrains is offering a solution with Junie Local, a coding assistant that trades raw power for total privacy.
Marcus: And finally, that Akamai report shows how your most enthusiastic employees could be creating huge security risks by using un-sanctioned AI tools.
Ivy: Before we go, Marcus, did you see that new AI-designed potato chip flavor?
Marcus: I did! It's like the AI just mainlined the entire Frito-Lay catalog. Human testers described the flavor as 'aggressive' and 'confusingly delicious'.
Ivy: So, the perfect snack to eat while contemplating how AI is automating our biases and leaking our corporate secrets. It fits.
Marcus: And that's our show for Tuesday, August 25th. Join us again tomorrow.
Ivy: Until then, try not to paste any state secrets into your AI. Or at least use a sanctioned one.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.