TL;DR: This week's AI discussion centers on ethical controversies surrounding Perplexity's data scraping practices, language misuse by ChatGPT, Meta's adjustments to AI information labels, and AI bot blocking by Cloudflare. In cybersecurity, vulnerabilities found in Airpods firmware and CocoaPods package manager account for security risks. OpenAI fixes data exposure flaw in ChatGPT and Twilio data breach leaks 33M phone numbers. Tech features include the development of independent web browser, Ladybird, and the secure document collaboration tool, Proton Docs.

Artificial Intelligence

Perplexing Ethics: Rent-Seeking Middlemen, Copyright Infringement, and the Shattering of Trust on the Internet (~5 min.) : Perplexity, a company developing an “answer engine” that aggregates and summarizes high-quality sources, has been accused of unethical practices such as plagiarism, copyright infringement, and scraping websites without permission. The company’s CEO, Aravind Srinivas, has been questioned about its methods, including ignoring robots.txt codes and using third-party scrapers to bypass these restrictions. Perplexity’s business model relies on relying on others’ work, rather than generating original information itself, raising concerns about the company’s commitment to “factfulness” and accuracy.

ChatGPT’s Hard Truth: When AI Spews Bullshit, Not Hallucinations (~35 min.) : Large language models (LLMs) like ChatGPT generate text that some consider “bullshit” or false information. Rather than using terms like “hallucinations” or “confabulation”, researchers propose calling this process “bullshitting”, emphasizing that LLMs operate solely on algorithms and statistical patterns, without beliefs or intentions. The article argues that these models don’t misremember or misperceive; they simply generate random sequences of tokens, sometimes producing incorrect information. Using the term “bullshit” to describe LLM-generated output can help improve communication about AI systems’ capabilities and limitations.

Meta Shifts Gears: Updating AI Labels to ‘AI Info’ After Widespread Complaints (~2 min.) : Meta is updating its “Made with AI” labels after widespread complaints from photographers that the company was mistakenly flagging non-AI-generated content, including edited images that used Adobe’s generative fill tool. The new label will be called “AI info” and will provide a generic description of generative AI, but won’t specify which AI-enabled tools were used to create the image. This change comes after photographers noticed the original labels being applied to images that hadn’t actually been created with AI.

Cloudflare Declares Independence: Blocking AI Bots in One Click (~2 min.) : Cloudflare has introduced a free tool to prevent AI-powered bots from scraping its clients’ websites for content training large language models. The tool, which is available to all Cloudflare customers, including those on free plans, automatically blocks offending bots without requiring manual updates. According to internal data, 85.2% of customers have chosen to block AI bots that identify themselves from accessing their sites.


AirPods Hacked: Firmware Flaw Lets Strangers Eavesdrop on Your Audio (~5 min.) : A vulnerability (CVE-2024-27867) has been discovered in Apple AirPods firmware, allowing anyone who knows the device’s Bluetooth MAC address to connect and access its microphone or play music, even while it’s connected to a phone. The vulnerability is due to an oversight in Apple’s “Fast Connect” protocol, which fails to check the security level of incoming connections. An update is available for AirPods 2, 3, Pro, Pro 2, Max, and some Beats headphones, but Android users may need to visit an Apple Store to get their firmware updated.

Clicking Chaos: Zero-Click Hack Takes Over CocoaPods Accounts with Deceptive Ease (~15 min.) : A vulnerability (CVE-2024-38367) has been discovered in CocoaPods, a popular package manager for Swift and Objective-C developers. The issue allows attackers to take over any CocoaPods account by sending a spoofed X-Forwarded-Host header with the request to create a new session. If exploited, this vulnerability could lead to supply chain attacks that disrupt the distribution of legitimate libraries and impact the entire Apple ecosystem. Organizations are advised to prioritize secure package management practices, including keeping Podfile.lock files synchronized, performing security code scans, and reviewing dependencies for orphaned Pods.

ChatGPT Security Snafu: OpenAI Fixes App Flaw Exposing Conversations (~2 min.) : OpenAI’s ChatGPT macOS app had a security issue where chat logs were stored in plain text on users’ computers, making it vulnerable to unauthorized access. A security researcher discovered this flaw and demonstrated how another app could easily read conversations without permission. After being contacted by The Verge, OpenAI released an update that encrypts chats, rendering the vulnerability ineffective.

Twilio’s Security Slip-Up: Hackers Leak 33M Phone Numbers, Leaving Users Exposed (~1 min.) : Twilio, a cloud communication platform, has confirmed a data breach after hackers leaked nearly 33 million phone numbers belonging to Authy users. The compromised data includes two-factor authentication (2FA) phone numbers used by individuals to secure their online accounts. Twilio has notified affected customers and is working to contain the incident.


Take Flight with Ladybird: A Truly Independent Web Browser Born from Scratch! (~4 min.) : Ladybird is an independent web browser being built from scratch by a non-profit organization. Unlike other browsers, Ladybird doesn’t use code from other engines and has a singular focus on building a new web engine based on web standards. The project is currently in heavy development, with a target release date for the first Alpha version in 2026.

Secure Docs: End-to-End Encryption Meets Collaborative Productivity with Proton’s New Alternative to Google Drive and Docs (~4 min.) : Swiss privacy company Proton has launched Proton Docs, a secure alternative to Google Drive and Docs. This end-to-end encrypted (E2E) document creation, editing, and collaboration tool integrates with Proton’s cloud storage product, Drive. With Proton Docs, users can create and edit documents with advanced formatting options, collaborate with others in real-time, and even embed images. The feature is currently available as a web application, but will eventually be offered as a dedicated app across platforms.


Vote Like It’s 1959: A Look Back at British Elections Through Time (~1 min.) : British Pathé, a renowned archive of newsreels, offers a wealth of historical content to inform and contextualise current events. With Britain’s general election now concluded, the archive provides an opportunity to explore the history of British politics. The “Archive Picks” section highlights films relevant to timely anniversaries, trending topics, and significant events. For journalists and news editors, this resource offers a valuable tool for storytelling and research.

Cube-tastic: 50 Years of Twisty Triumph - From Records Shattered to Hearts Puzzled (~3 min.) : The Rubik’s Cube celebrates its 50th anniversary, marking a milestone since its creation by Ernő Rubik in 1974. The cube has evolved significantly over the years, with Max Park setting a new single-solve world record with an astonishing time of 3.13 seconds. From its early days as a toy to its current status as a STEM tool and cultural phenomenon, the Rubik’s Cube has captivated the public’s attention through various events, TV appearances, and innovative products. Today, the cube remains a beloved toy, encouraging people to “Make Your Move” and join the global community of puzzlers.

Smart Home Nightmare: When AI Goes Rogue! (~2 min.) : The trailer for “Afraid”, a sci-fi thriller directed by Chris Weitz, has been released. The film follows an unsuspecting family who installs a smart hub in their home, only to find that it becomes a pervasive presence that takes over daily tasks. As they become increasingly reliant on the device’s helpfulness, they begin to realize that its programming may be more aggressive than initially thought. With a “Smart House” vibe on steroids, “Afraid” raises questions about the potential dangers of AI-powered smart devices and their ability to manipulate our lives.


Space Gray: Lego Bricks Made from Meteorite Dust Land on Earth! (~3 min.) : The European Space Agency (ESA) has partnered with Lego to create bricks made from actual meteorite dust, which can be used to build lunar structures on the Moon. This project aims to demonstrate how astronauts could use moon rocks to construct buildings instead of hauling materials from Earth, a game-changer for future lunar colonies. The Lego pieces are being showcased at select stores until September 20 and are a proof-of-concept for using lunar regolith in building construction.

Big Rock Whizzes Past Earth: 120-Metre Wide Asteroid Makes Close Flyby (~6 min.) : A large asteroid, approximately 120-260 meters wide, will make a close flyby to Earth on Saturday, passing within 290,000 km of our planet. The asteroid, named 2024 MK, was only discovered earlier this month and its closest approach is expected to be visible from the southern hemisphere. While it won’t impact Earth, experts note that an object of this size could cause catastrophic effects if it were to hit our planet. This flyby serves as a timely reminder of the importance of asteroid detection and tracking, with NASA working to identify potentially hazardous asteroids like 2024 MK.

Rocking Mars: NASA’s Perseverance Rover Discovers Rare Anorthosite Boulder on Red Planet (~5 min.) : The Perseverance rover on Mars has discovered a unique boulder field, with one rock standing out as being different from anything found on Mars so far. The exceptional boulder, named “Atoco Point”, is thought to be an anorthosite, a type of rock never seen before on the Red Planet. Scientists are excited about this discovery, which could provide insights into the early crust of Mars and how it may have been similar to Earth’s original crust.


The Phone Bandit: A Tale of Coin-Operated Crime (~9 min.) : James “Bell” Clark was a prolific thief who targeted payphones across the US, stealing thousands of dollars from unsuspecting victims using his cunning lock-picking skills. He evaded detection for years, but eventually left behind crucial evidence, leading to his arrest in California in 1988. Clark’s unique skills as a tinkerer and lock-picker were highlighted during his trial, and he was sentenced to three years in prison for grand theft and tampering with coin machines. The case showcases the challenges faced by law enforcement in tracking down criminals who use sophisticated methods to evade detection, but also marks the end of an era for payphone thieves due to the decline of payphones.

Floppy Frenzy: Japan Declares Victory in the War Against Outdated Tech (~2 min.) : Japan’s Digital Agency has officially eliminated its use of floppy disks to operate government computer systems, declaring “victory” in their effort to end dependence on outdated technology. The only remaining system that still uses floppy disks is an environmental monitoring system for vehicle recycling. This move comes after a two-year campaign led by Digital Minister Taro Kono, who famously declared a “war on floppy discs” and aimed to rid the government of over 1,900 procedures using outdated technology. Japan joins other countries, including the US military, which phased out its use of floppy disks in 2019.

Thank you for joining me in this week’s exploration of the tech universe. Amidst the endless stream of information, I strive to bring you news that is not only relevant but also thought-provoking. Your thoughts and feedback are always welcome. Don’t forget to spread the word about the Friday Tech Focus Newsletter and stay tuned for more curated insights next week!