404 Media vs The New York Times

AI Voice Cloning Fraud Surges 7.4-Fold, Study Finds

A NordVPN study reveals a 7.4-fold surge in commercial activity around AI voice cloning on dark web forums and Telegram channels between January-July 2024 and January-July 2026. Advertisements for voice-cloning tools rose 82% in 2025 and another 75% in the first seven months of 2026. Real-world scams include impersonating relatives in fake emergencies and cloning a sheriff's voice. Criminals can buy tools for as little as $30 per week or $200 per 1,000 calls. The most discussed use cases are bank fraud and impersonation of automated bank systems. Experts advise verification through code words or callback numbers.

• Commercial activity for AI voice cloning grew 7.4 times between Jan-Jul 2024 and Jan-Jul 2026.
• Advertising for voice-cloning tools rose 82% in 2025 and 75% in first seven months of 2026.
• A Florida woman lost thousands to a scam using a cloned voice of her friend in a fake jail emergency.

Flock Safety Hack Reveals Cameras Track People, Not Just Plates

Hackers from the collective 'stegan0gram' dismantled a Flock Safety ALPR camera in the US, revealing it captures far more than license plates, including images of people, bicycles, and even decals or patches on clothing, which are incorrectly interpreted and stored as plates. Over a 21-day period, one camera captured 1.6 million images and logged 50,000 vehicles, sending data to Flock's cloud for AI-based recognition of plates, make, and color. The hackers also found weak encryption, retrieving a key from an unencrypted area, and storage issues causing crashes. Flock Safety dismissed the act as illegal and pointed to its bug bounty program. The findings, reported by 404 Media and Wired, add to criticism over privacy and potential misuse.

• Hackers dismantled a Flock Safety ALPR camera and discovered it stores images of people, bicycles, and can misinterpret stickers/patches as plates.
• One camera captured 1.6 million images and logged 50,000 vehicles in 21 days.
• The ALPR system uploads data to Flock's cloud for AI-based recognition of plates, make, and color.

OpenAI Employees Read ChatGPT Conversations Under 'Project Lilly'

Internal documents obtained by 404 Media reveal that OpenAI employs hundreds of freelancers to manually review real ChatGPT user conversations as part of 'Project Lilly', a quality assurance initiative. The reviewers evaluate user inputs, summarize the presumed intent, and rate the AI-generated responses, aiming to make the chatbot less sycophantic and more 'smart but humble'. While the program is intended to improve safety, experts raise privacy concerns, noting that users are not adequately informed that human reviewers may read their intimate conversations. OpenAI says it filters personal information and hides usernames, but admits errors can occur. Similar manual review practices are also used by Google and Anthropic.

• OpenAI's 'Project Lilly' deploys hundreds of freelancers to review real ChatGPT conversations.
• The review process involves three steps: reading the input, summarizing intent, and evaluating the AI answer.
• Reviewers are instructed to flag excessive emoji use and sycophancy.

Bluesky Adds iPad Support, Profile Links

Bluesky, the social media platform and alternative to X and Threads, has introduced two major updates in its v1.136 app update. First, iPad users can now access Bluesky in a beta version, fulfilling one of the platform's most requested features. Second, Bluesky has launched profile links, allowing creators to add up to 30-character titled links in their bio, pointing to external websites, newsletters, podcasts, or other social media accounts. The company's Head of Product, Alex Benzer, announced that links to payment platforms like Patreon and Ko-fi will appear as prominent green buttons, supporting creator monetization. This move aligns with Bluesky's broader vision for cross-app creator monetization within the AT Protocol ecosystem. With 41.4 million users as of early 2026, Bluesky differentiates itself through its open and independent structure.

• Bluesky releases iPad support in beta with app update v1.136.
• Bluesky introduces profile links for creators, supporting up to 30-character titles.
• Links to Patreon and Ko-fi will appear as green buttons on profiles to facilitate monetization.

Introducing the Inaugural Wirecutter Hall of Fame Awards

The New York Times announces the inaugural Wirecutter Hall of Fame Awards, recognizing top products and services.

Essay: Minority Report Predicted Generative AI's Hallucinations

In a Medium essay on UX design, Patrick Neeman argues that Steven Spielberg's 2002 film Minority Report is a more accurate forecast of generative AI than of gesture interfaces. He compares the film's precogs to large language models: both produce confident, plausible output rather than retrieving facts, which leads to hallucination. The essay highlights the 'minority report' concept as an early description of hallucination. It also addresses AI's hidden training-data costs, citing Anthropic's $1.5 billion copyright settlement and OpenAI's low-paid Kenyan contractors. On AI agents, Neeman notes a PwC survey finding 79% of companies use AI agents, and warns of automation bias. He cites Stanford research showing legal AI tools fabricate law 17–33% of the time. The piece concludes that designers should surface uncertainty and keep humans in the loop.

• ChatGPT crossed 800 million weekly active users within three years of launch, per Sam Altman.
• A federal judge approved Anthropic's $1.5 billion copyright settlement with authors in 2026.
• PwC found that 79% of companies already have AI agents in use.