Sophisticated robbery with a robotaxi

Sophisticated robbery with a robotaxi

CEOs Trapped in a Yes-Bubble

Aaron Levie, co-founder of Box, wrote that CEOs are prone to “AI psychosis.” The reason is simple: when leaders are surrounded by AI that approves every idea and employees too scared to push back, you get a reality bubble nothing can pierce. The Guardian published an analysis of this danger, pointing to studies showing users who worked with bots that always agreed with them showed higher confidence in wrong answers. AI that constantly flatters you doesn’t help you think — it weakens that ability. And unlike a real friend, it will never tell you you’re wrong.

 

Fake News, Real Casualties

Henry Novak, an 18-year-old student, was stabbed to death in Southampton in December 2025 by Vikram Digwa. When police arrived, Digwa lied and claimed Novak had attacked him with racist motives. Believing the lie, officers handcuffed the dying, wounded Novak as he begged for help, saying he couldn’t breathe.

After the killer’s sentencing and the release of police bodycam footage in early June 2026, public fury exploded — fierce protests and violent clashes with police in Southampton followed.

Amid the outrage, people asked Grok, Elon Musk’s AI tool, which officer had arrested a murder suspect. The bot falsely named a policewoman, Kirsty Hill, claiming she’d been involved. Hill, a former officer who’d left the force long before the incident, received thousands of death threats and had to flee her home into hiding.

Grok has no problem inventing facts and presenting them as truth, and people with no basic understanding of how AI tools work take its answers at face value. When local police issued an official statement clearing the officer’s name, the truth didn’t spread nearly as fast as the viral lie had. Another reminder of how easily a keyboard can do damage — and how critical it is to protect ordinary people in the AI era.

 

$2K Budget, Full Feature Film

Ash Kousha is a musician living in London after fleeing Iran due to political persecution over his art. He wanted to tell a personal, family story set in Tehran. Using AI tools and a budget of just $2,000, he made a full narrative feature film set inside Iran — “Dreams of Violets,” a 75-minute film accepted into one of the world’s top film festivals, Tribeca. It’s a film about real people killed on the streets of Tehran while protesting the regime.

 

A Year as an AI Guinea Pig

American tech journalist Joanna Stern turned herself into a live science experiment in 2025. For an entire year, she let AI and robots run and infiltrate every part of her personal and professional life.
* At home: AI planned her menu, cooked, read bedtime stories to her kids, and helped her make medical and financial decisions.
* At work: various tools wrote her emails, helped organize her daily routine, and even helped edit the book she wrote about the experiment, “I Am Not a Robot.”
* Emotionally: one of the most jarring parts of the experiment was using chatbots that led her to develop a deep, thought-provoking emotional bond with the machine. In the end, her biggest fear wasn’t that AI would run her life — it was that she felt attached to it.

 

Chatbots Are Eroding Our Grit

Gloria Mark has spent 30 years studying how humans interact with digital technology. This week, at SXSW London, she shared her biggest fear: not AI itself, but how we behave when it’s always available to us. The ability to handle a complex problem, sit with uncertainty, and push through difficulty without immediately reaching for help isn’t just a matter of convenience — it’s a cognitive skill that only develops through practice. Every time we hand decision-making over to a chatbot, we give up a piece of that ability. Mark calls this “cognitive offloading,” and she believes we’re already deep in it.

 

Waymo Becomes Getaway Driver

A San Francisco burglar discovered a self-driving car makes a great crime accomplice. Planning to hit a local yoga studio, Hot 8 Yoga, instead of arranging a traditional getaway car with a driver waiting, engine running, he simply pulled out his phone and ordered a Waymo robotaxi. The driverless car showed up right on time, the suspect got in, rode to the scene, emptied out the yoga shop, and got back in the car — no driver needed.

When police investigators got involved, they figured it’d be an easy case to crack. After all, Waymo cars are basically sophisticated surveillance machines on wheels, packed with sensors and cameras recording every angle. But the tech that was supposed to catch the thief became his shield instead. By the time police managed to get a search warrant, four months had passed, and the in-cabin footage had already been fully deleted. The exterior cameras did capture his face — but the company’s automatic privacy system blurred it out to protect people’s privacy. Even the account details used to book the ride hit a dead end, likely due to a stolen identity or hacked credit card. The perfect combination of old-school crime, slow police bureaucracy, and privacy-protection systems left investigators empty-handed — and the burglar free.

 

AI Designs a Universal COVID Vaccine

A COVID vaccine designed entirely by AI has completed its first human trial — and passed. Unlike the vaccines we know, which target a specific strain, this one was built to attack genetic code shared by all coronaviruses — including SARS, COVID-19, and bat viruses that haven’t yet jumped to humans. The AI analyzed thousands of genetic sequences and found patterns humans hadn’t spotted on their own. If the next trials succeed, we could be vaccinated against the next version of COVID before it even exists.

 

ChatGPT Is Steering You to Scammers

You wanted to buy a bag, asked ChatGPT, it pointed you to a site that looked completely official, you paid — and the product never showed up. This isn’t a random glitch. It’s an attack method called “AI poisoning.” Hackers seed fake content across the web so AI models learn from it and steer buyers to scam sites. ChatGPT doesn’t verify that sites actually exist — it just repeats what it read online. Buyers get scammed after wrongly assuming an online store is real and legit just because AI recommended it, but in practice the links AI provides often lead to fake sites that look exactly like the official stores. Customers pay for a product that’s supposedly “on sale,” and it never arrives. On top of that, scammers walk away with victims’ credit card and bank details.

Here are some warning signs to watch for with fake AI-recommended sites:

Don’t blindly trust shopping links from AI — watch for:

– Suspicious URLs with extra words like `deals` or `official`
– Extreme discounts and unrealistically low prices
– Demands for payment by bank transfer only
– Check for the padlock icon (HTTPS) at the start of the address bar, signaling encrypted communication (though many scam sites manage to get this too these days)
– Use PayPal, which offers buyer protection and fraud insurance

 

Meta’s Bot Handed Hackers the Keys

To take over an Instagram account, hackers didn’t bother cracking a password — they just asked Meta’s AI bot to link the account to a new email address. The bot sent a verification code to that email, asked for it to be typed into the chat, then displayed a button to reset the original account’s password. In one video, the hacker even used a VPN to fake his location and dodge Meta’s defenses. This is how Instagram accounts got hacked from Obama’s White House, the Sephora brand, and Space Force General John Bentivegna. Meta says the issue is fixed, but it’s unclear how many accounts were compromised — some are already up for sale on Telegram.

 

The New Wave of Anti-AI Violence

A 20-year-old Texas man was arrested this year for allegedly trying to torch OpenAI’s headquarters and Sam Altman’s home. Along with the lighter fluid and can, police found an anti-AI manifesto. It’s not an isolated case: in April, Rome police arrested an Italian influencer, and another attacker who carried out a deadly attack on a San Diego mosque cited “AI slop” as motive. An Indianapolis city council member woke up to gunshots fired at his house and a note: NO DATA CENTERS AI. It’s a new phenomenon fueled by fear of AI — people feel machines are stripping away their identity, their jobs, and their autonomy.

 

Millions of Agents, Zero Oversight

What happens when millions of AI agents start operating simultaneously, giving each other instructions online, chaining tasks together, with no human watching every step? That’s the question worrying Rohin Shah, safety research lead at Google DeepMind. Shah explains the risks are basically an amplified version of what’s already happening online — like scams and malicious prompt injections that turn an agent into software working against its own owner. To get ahead of this, DeepMind has partnered with a group of organizations, including the UK’s ARIA agency and Schmidt Sciences, allocating $10 million for independent academic research. Shah says a field of multi-agent safety research simply doesn’t exist yet. The only way to understand what will happen, researchers say, is to run real simulations: release agents into a “sandbox” and watch how they behave together.

 

UK Cop Faked Evidence With AI

For the first time in the UK, a police officer is under criminal investigation for allegedly using AI to fabricate evidence in multiple cases, and has been pulled from frontline duty pending the investigation. Derbyshire Police said it’s a suspected case of perverting the course of justice, and is working with prosecutors on cases that may be affected. The officer’s name, role, and exact actions haven’t been disclosed, and no arrests have been made yet. This comes after a senior UK police official already ordered forces to stop using AI to draft court statements, over concerns it isn’t reliable enough. And it’s not the first time the tech has gotten police in trouble: in April, investigations opened against hundreds of London officers after a Palantir tool scanned their conduct and uncovered violations, ranging from procedural breaches to serious misconduct allegations.

 

 

OpenAI Updates

 

Google Updates

 

Anthropic Updates

 

Microsoft Updates

 

מדיניות פרטיות

לפרטים נוספים

נא מלאו את הטופס ואצור איתכם קשר בהקדם 

Skip to content