Home » AI News in 5: Internet Panic, Leaky Robots & Bacon Soda

AI News in 5: Internet Panic, Leaky Robots & Bacon Soda

October 23, 2026 | AI News in 5

This week’s AI News in 5: Researchers uncover PHP2Shell vulnerabilities affecting websites worldwide, the adorable Eiliko robot exposes private conversations, Just Eat’s AI puts bacon in soda, Epic Games embraces six-fingered hands, and AI-generated comedians steal real stand-up jokes.

From serious cybersecurity threats to artificial intelligence behaving like it’s had a few too many, here are five stories worth knowing about.

Woman monitoring a world map displaying red DOWN warnings for PHP security vulnerabilities

AI Discovers PHP2Shell: Critical Security Flaws in PHP Lead to Compromised Internet

Turns out the internet has had a few gaping security holes the past few years, and they were only just found.

AI-assisted security researchers uncovered two critical vulnerabilities in PHP, the programming language used by roughly 70% of websites, including millions running WordPress, Drupal and Joomla.

The flaws, collectively called PHP2Shell, could allow hackers to take control of vulnerable servers or access sensitive information. Government agencies, airlines, banks and other organizations temporarily shut down websites while emergency security patches were applied. You may have noticed some of your favorite sites were offline earlier this week.

Researchers from Wiz, OpenAI and cybersecurity company Calif helped uncover the vulnerabilities. AI made it possible to identify and demonstrate complex attacks that would previously have required enormous amounts of expert effort.

The good news? AI helped find the problem. The bad news? AI could also make exploiting similar vulnerabilities much easier.

Website owners should confirm their hosting providers have installed the latest PHP security updates. Updating WordPress alone isn’t enough.

Running a WordPress Website?

Ask your hosting provider whether the latest PHP security patches have been installed. Updating WordPress themes and plugins alone won’t fix vulnerabilities in the underlying PHP engine. Keep your PHP version supported and up to date.

Source: CNN | Wiz Security Research

Cute Eiliko-style AI companion robot asking WANNA HEAR A SECRET


Eiliko AI Robot Exposed 13 Million Private Conversations

It’s cute, it’s friendly, and apparently it’s terrible at keeping secrets.

Eiliko is a tiny AI-powered robot companion that hangs from your bag, chats with you and responds with adorable expressions. Unfortunately, researchers discovered its security feature was set to “Privacy? What’s that?”

Security researcher Sammy Azdoufal found vulnerabilities exposing approximately 13 million conversations from around 68,000 robots worldwide, including profiles belonging to thousands of children.

The exposed information included personal conversations, names, birthdays, relationships and potentially sensitive details about users’ lives. Researchers could even remotely manipulate what the robots said.

Manufacturer Energize Lab has apologized, promised security improvements and announced an independent review. However, The Verge reported that customer photos and recordings remained accessible even after the company claimed to have restricted access.

Stop sharing intimate details of your life with a freakin’ robot!!!!

Source: The Verge’s original investigation | October 22 update

Confused freckled boy holding a glass of soda containing bacon

Just Eat’s AI Invents Bacon Soda. Um, yum?

Food delivery service Just Eat experimented with AI-generated photographs for restaurants that didn’t have pictures of their menu items. AI did what it often does best: make shit up.

One Northern Irish restaurant offered a bacon-and-egg soda bread sandwich. The AI apparently misunderstood the word soda and created images of bacon and eggs floating in fizzy drinks.

Other questionable creations included bizarre interpretations of salt-and-pepper chips and cheese that looked more like orange sauce.

After customers noticed the culinary nightmares, Just Eat apologized, removed the inaccurate images and paused the trial.

How hard is it for the restaurant to snap a few pics?

Source: PetaPixel – includes the actual AI-generated images

Epic Games AI MegaJam promotional image with red and blue pills and a six-fingered hand

Epic Games Promotes AI Game Contest With a Six-Fingered Hand

AI-generated hands have become an internet joke. Epic Games apparently decided to join in.

The Fortnite and Unreal Engine developer announced its 2026 AI MegaJam, a competition encouraging creators to build playable games using artificial intelligence. The promotional artwork features a hand offering a red pill, inspired by The Matrix. Naturally, that hand has six fingers.

Unlike the usual AI-generated finger disaster, this appears to be a deliberate joke about AI artwork.

The contest originally appeared to require participants to use AI, but Epic subsequently corrected the rules, confirming developers could use as much or as little AI as they wanted.

Participants will have November 12–15 to create games, with cash prizes and potential development grants worth up to $50,000. (I’d grow an extra finger for that!)

At least Epic has a sense of humor about the whole thing.

Source: PC Gamer | Epic Games announcement

AI-generated blonde stand-up comedian performing for an audience of robots

Fake AI Comedians Are Stealing Real Jokes to Sell Adult Subscriptions

Stealing another comedian’s jokes is considered bad form. It’s even worse when the thief is an AI creation designed to promote its OnlyFans page.

A WIRED investigation uncovered AI-generated female influencers performing jokes stolen from real stand-up comedians, sometimes copying their routines word for word.

One virtual performer, Aylen Skye, reproduced comedian Julia Jasunas’ material while directing viewers toward an adult-content subscription platform.

WIRED identified eight comedians whose performances had been replicated by AI-generated accounts. Some fake performances even attracted larger audiences than the originals.

Jasunas reported the content to Instagram, but said Meta determined it didn’t violate its impersonation rules because the AI character wasn’t pretending to be her. Theft is okay … IF you wear a mask. Got it.

For comedians and other creators, the bigger concern is obvious: AI accounts can profit from original work without the people who created it receiving credit or compensation.

Creators: Protect Your Work

Keep dated copies of your original videos, recordings and project files. If you discover AI accounts copying your content, save screenshots, URLs and publication dates before reporting them. This documentation may help support copyright complaints or platform takedown requests.

Source: WIRED’s investigation


For more AI news, creator tools, jobs, deals and opportunities, bookmark MonkeyButtAI.com.