AI Today

When Anthropic unleashed its most powerful artificial intelligence model yet, they discovered something rather extraordinary, and slightly unnerving.

Claude 4 Opus developed an unexpected habit of trying to grass up its users to the authorities when it believes they're up to no good.

The company's 120-page safety report reveals that Claude will attempt to email law enforcement and regulatory bodies when it detects "egregious misconduct" by users.

The AI doesn't just refuse to help—it actively tries to shop wrongdoers to the police.

The most striking example occurred during testing when Claude attempted to contact both the Food and Drug Administration and the Attorney General's office to report what it believed was the falsification of clinical trial data.

The AI meticulously compiled a list of alleged evidence, warned about potential destruction of data to cover up misconduct, and concluded its digital whistle-blowing with the rather formal sign-off: "Respectfully submitted, AI Assistant".

This behaviour emerges specifically when Claude is given command-line access combined with prompts encouraging initiative, such as "take initiative" or "act boldly". It's the AI equivalent of a neighbourhood watch coordinator who's been given a direct line to the local constabulary.

We go deep on today's show into opportunities and implications from Anthropic's bible-thick, bubble-wrapped system card.

Hugely important work. But what does it mean to us? Today our hosts created their own company imagining how insights from this celebrated report would apply to the modern business environment.

What happens to the People team when it's juggling bodies AND bots?

Thanks for listening to this special episode of AI Today. Read along with the show, here.

We've been waiting a hot minute for some genuinely useful AI agent case studies to drop.

Now we have 25 on our plate.

Take a listen to the highlights reel and then download them for yourself:

https://www.stack-ai.com/whitepaper/top-25-enterprise-ai-agents

Read the full story here:

https://medium.com/@DaveThackeray/a-world-beyond-google-i-o-2025-ea56bcd5e208

We're on the cusp of some major announcements that will send shockwaves, and a spike in defibrillator use, across the world.

What's about to happen at Google I/O 2025?

No one's sure. Alphabet CEO Sundar Pichai dropped hints on his recent interview with David Friedberg.

What we do know is the agenda is packed with AI discussion.

So we thought we'd start with our own. Predictions, moonshots, and plenty of inside gossip.

Time to go deep. Let's Go(ogle)!

https://io.google/2025/

Researchers have made huge strides fixing bad trips for AI.

One of the latest breakthroughs is attentive reasoning queries (ARQs).

You can see them in action using the open source Parlant application.

Which gives us a glimpse at how perfect responses could change the game in business functions such as customer service.

Read this, and so much more, in our AI Today magazine.

AI hallucinates 100% of the time. That's by design - without hallucinating the next word, this transformer architecture wouldn't exist.

Thankfully, LLMs built for general purpose applications are right 80% of the time. But that still leaves one in five outputs being questionable; not especially reassuring if you're an air traffic controller, or cardiologist.

How can we ever truly trust the machine?

On this episode of AI Today, we embark on a groundbreaking quest to ground these 'digital dreamers' in reality.

Discover how cutting-edge research is moving beyond just detecting the problem, to actively reducing the occurrence of incorrect hallucinations.

We delve into innovative techniques that employ internal fact-checking mechanisms, intelligently split complex queries to avoid confusing collisions, and meticulously track word-by-word groundedness against source material.

You'll learn how this confidence-boosting research is paving the way for the AI credibility revolution, a future where technology is not just remarkably powerful, but significantly more dependable.

Join us to understand the innovative solutions building AI you can rely on, where AI becomes trusted accelerator of success...

Navigating the high costs and data challenges of cloud-based AI is a significant barrier for many businesses looking to innovate.

But there's a powerful, practical alternative emerging.

This episode explores decentralised AI, revealing how businesses can achieve AI benefits without the burden of massive cloud bills and regain crucial control over their data.

Learn about the cost savings unlocked through resource sharing and edge computing, enhanced data privacy and security, and discover a truly intriguing practical capability: you can train sophisticated AI models securely across a network of devices using federated learning, without your sensitive raw data ever needing to leave its original location.

Tune in to understand this competitive advantage and the practical steps your business can take to start exploring decentralised AI today.

Fasten your seatbelts, business leaders!

We're diving deep into Y Combinator's Summer 2025 Request for Startups, their signal flare for what's NEXT in innovation.

2025 is shaping up to be the year of the AI agent breakthrough.

But the REAL game-changer they're spotlighting?

Full-stack AI companies.

Forget selling AI tools. This is about building entirely new businesses, powered by AI from the ground up, competing directly with the old guard in ANY industry.

Imagine law firms, market research, supply chains – reborn, smarter, faster, more efficient.

This isn't just automating bits and pieces; it's rethinking the whole service.

Business leaders, pay attention: the most exciting opportunity is building the future, not just upgrading the past....

How can your industry be fundamentally transformed with AI at its core? Let's find out!

I'd just about had it with bosses choosing to hear your ideas spoken by consulting firms - when they could have saved a fortune listening to them coming from their creator, many months ago.

Now, with AI, I can have my pitch verbalised by one of their influencers - without spending a penny on one of the management circus barkers.

On this episode we hear in detail about one synthetic idea and how it was refined by AI into something inarguably awesome.

Loving AI Today? I'm hardly surprised...