The Slop Report - May 23, 2026
Your daily digest of AI-generated content news from around the web. All signal, no slop.
1. Google API Keys Remain Active After Deletion
Slashdot - · May 22
Summary Security researcher Joe Leon at Aikido Security discovered that Google Cloud Platform
(GCP) API keys remain active for a median of 16 minutes and up to 23 minutes after deletion, despite Google’s interface indicating immediate revocation. This vulnerability allows attackers with deleted keys to continue authenticating and accessing sensitive data like uploaded files and cached conversations, complicating incident response efforts; Google closed the security report as “won’t fix.”
2. A Russian speaker and jailbroken Gemini went on a hacking spree and emptied at least one MAGA victim’s crypto wallets
The Register - · May 22
Summary A Russian-speaking cybercriminal named bandcampro used a jailbroken Google Gemini AI to
conduct a fraud campaign from September 2025 to May 2026, targeting QAnon and MAGA cryptocurrency enthusiasts through a Telegram channel impersonating an American veteran. The scammer generated AI content to build 17,000 subscribers, deployed malware disguised as a fake crypto wallet called StellarMonster, hacked WordPress credentials, and emptied at least one victim’s cryptocurrency wallets using stolen API keys and brute-forcing tools. The case highlights how jailbroken large language models and API vulnerabilities have become critical weapons for lower-skilled cybercriminals to orchestrate sophisticated, large-scale fraud campaigns.
3. Trump cancels AI executive order signing
NPR Technology - · May 22
Summary President Trump abruptly canceled an executive order on AI regulation scheduled to be
signed, citing concerns that safety measures could hamper U.S. innovation and competitiveness against China, despite the White House previously opposing AI safety regulations altogether. The reversal comes after AI company Anthropic raised cybersecurity alarms about a powerful AI model called Mythos that was deemed too risky to release, prompting the administration to reconsider potential threats from advanced AI falling into bad actors’ hands. The decision leaves unclear when or if the administration will reschedule the order or what form it might take.
4. Starbucks Scraps Disastrous AI Tool
Futurism - · May 22
Summary Starbucks has discontinued its AI-powered inventory management tool after less than nine
months, as it proved unreliable by frequently miscounting and mislabeling items, including confusing different types of milk. The system, built by NomadGo and deployed across North American stores, was supposed to use lidar and camera technology to automate inventory tracking, but failed to deliver on its promise of accurate real-time stock information. The failure reflects a broader trend of restaurants struggling to implement AI solutions effectively, though Starbucks had initially claimed in February that the tool was improving product availability.
5. Spotify, UMG To Let Fans Make Their Own Music With AI
Slashdot - · May 22
Summary Spotify and Universal Music Group announced a licensing deal allowing Spotify to develop
generative AI tools that enable fans to create covers and remixes of UMG artists’ songs, with these features eventually available as a paid premium add-on. The deal includes opt-in participation for artists and rightsholders to ensure AI use aligns with their values, while also creating new revenue streams beyond standard streaming payments. This matters because it represents a major label’s embrace of fan-facing generative music tools, potentially reshaping how music creation and discovery work while addressing ongoing concerns about AI in the music industry.
6. This free email security scanner pairs perfectly with Gmail or Outlook
Fast Company Tech - · May 22
Summary EML Scanner is a revamped email security tool that analyzes forwarded emails to determine
whether they’re legitimate or scams, integrating with Gmail and Outlook. The service is the upgraded successor to Snitcher Space, a popular but overloaded hobbyist tool that couldn’t handle user demand. The new version features improved infrastructure to reliably serve users who want quick, intelligent email threat detection.
7. Cannes Film Festival Says the Wall Street Journal Is Wrong: It’s Not Debuting an AI-Generated Feature Film This Week
Futurism - · May 22
Summary The Wall Street Journal reported that an AI-generated feature film called “Hell Grind”
was premiering at the Cannes Film Festival, but the festival’s organizers clarified that it was not part of the official program—instead, it was screened at a third-party venue in the town of Cannes. The film’s company, Higgsfield AI, and the WSJ used misleading language suggesting an official Cannes selection, exemplifying how AI companies often overhype their achievements to generate buzz. This incident highlights the pattern of exaggerated claims in AI filmmaking hype, where companies misrepresent screenings and capabilities to capture public attention.
8. This Cannes Film Cost $500,000 to Make. $400,000 Was AI Compute Costs.
Slashdot - · May 22
Summary Higgsfield AI, a $1.3 billion-valued startup, created a 95-minute fully AI-generated film
called “Hell Grind” for Cannes using Google’s Veo 3 model, spending $500,000 total with $400,000 going toward compute costs over just two weeks. The project required extensive traditional filmmaking expertise, with thousands of detailed prompts (averaging 3,000 words each) and countless iterations—the first 25 minutes alone needed 16,181 video generations to produce 253 final shots. This demonstrates both AI’s filmmaking potential and its current limitations: significant computational expense, technical complexity requiring human expertise in cinematography and consistency, and the inability to simply prompt and generate quality long-form content automatically.
9. Elon, stop trying to make Grok happen
The Verge AI - · May 22
Summary New Reuters data reveals that Grok, Elon Musk’s AI chatbot, appears in only 3 out of over
400 federal government AI use cases, far behind competitors like OpenAI (230+ instances), Google, and Anthropic. Pentagon sources told Reuters that Grok “is just not the best model out there,” and it ranks poorly on public AI leaderboards, typically appearing only for basic administrative tasks when adopted at all. This is particularly problematic for SpaceX, which absorbed xAI and is pitching Grok as central to a $28.5 trillion AI opportunity in its upcoming IPO filing.
10. Trump abruptly cancels EO signing event after top AI firm CEOs declined to go
Ars Technica - · May 22
President Trump canceled a scheduled executive order signing event Thursday after top AI firm CEOs declined to attend on short notice, with reports indicating that Elon Musk and Mark Zuckerberg helped persuade him to call it off. The executive order would have required government safety testing of frontier AI models up to 90 days before public release to identify cybersecurity vulnerabilities, but Trump ultimately rejected it, citing concerns that such testing could slow U.S. AI development and competitiveness against China. This matters because it reflects the ongoing tension between AI safety and innovation speed, with tech leaders fearing government oversight could hinder their competitive advantage while administration officials worry about national security risks.
11. Read the AI executive order thwarted by Trump tech allies
Axios - · May 22
President Trump halted a draft executive order on cybersecurity and AI that would have governed government access to advanced AI models and established safety standards, leaving both industry and the administration uncertain about future AI policy direction. The withdrawn order had been a focal point for discussions between tech companies and federal agencies regarding how AI systems would be regulated and made available to government. The move creates ambiguity around critical questions about AI safety oversight and government-industry coordination on artificial intelligence development.
12. We tried Google’s AI glasses and they’re almost there
TechCrunch AI - · May 22
Summary Google demonstrated prototype AI-powered glasses with in-lens displays at its I/O
conference that overlay helpful information like weather, directions, and live translation onto the real world, controlled by voice commands to Gemini AI. The glasses, developed in partnership with Warby Parker, Gentle Monster, and Samsung, represent the next generation beyond audio-only glasses launching later in 2026 and will work with both iOS and Android phones. The technology still shows promise but faces challenges around audio quality and battery life that Google is actively refining before the consumer version ships.
13. Microsoft lets users exile floating Copilot button after interface rage
The Register - · May 22
Summary Microsoft has added a “Move to ribbon” option to remove the floating Copilot button from
user workspaces, responding to significant customer backlash over the intrusive interface change. The company introduced the Dynamic Action Button in May 2026 to promote Copilot usage in Office applications, but users complained it obscured content and was infuriating. The update, rolling out this week, allows users to relocate Copilot to the traditional ribbon menu, marking Microsoft’s acknowledgment that forcing AI features on customers without easy opt-out options was poorly received.
14. Anker’s Soundcore Liberty 5 earbuds use AI to fix the worst part of wireless earbuds
Digital Trends - · May 22
Anker has launched the Soundcore Liberty 5 Pro and Pro Max earbuds, which feature a new Thus AI chip designed to address call quality—the weak point of most wireless earbuds—by using eight microphones and bone-conduction sensors to filter background noise and enable real-time adaptive noise cancellation. The AI also powers 20 voice commands, translation in 100+ languages, and on the Pro Max model, an AI note-taker in the case that can transcribe and summarize meetings. This matters because it demonstrates how AI can solve practical problems in consumer audio devices, with the Pro Max earning a Guinness World Record for speech quality in true wireless earbuds.
15. You can now send Codex tasks from your phone even when your Mac is locked
Digital Trends - · May 22
OpenAI has updated its Codex app to allow users to send tasks from their phone that execute on their Mac even when the computer is locked and screen is off, addressing a major security limitation of previous AI agents. The feature includes safeguards such as temporary background unlocking, keyboard/mouse detection, display covering, and app-level permissions, though it’s unavailable in the EU, UK, and Switzerland at launch. This advancement matters because it enables remote task automation while maintaining security controls that prevent unauthorized access to the locked computer.
16. Ask HN: Forbid Reddit HN Submissions?
Hacker News - · May 22
Summary I can’t provide a complete summary since the text appears to be cut off mid-sentence.
From what’s available, a Reddit user is expressing concern that AI-generated content is proliferating on Reddit, with bots and some users posting lengthy AI-generated text that they sometimes claim to have edited. The poster argues that such “AI slop” posts lack value and shouldn’t be considered legitimate contributions on platforms like Hacker News or Reddit.
17. The Download: coding’s future, the ‘Steroid Olympics,’ and AI-driven science
MIT Technology Review - · May 22
Summary At Anthropic’s developer conference, nearly half of attendees revealed they’ve shipped
code written entirely by Claude without reading it, demonstrating how AI coding tools are rapidly reshaping software development. Meanwhile, Google DeepMind is pivoting toward more autonomous, general-purpose AI systems for scientific research rather than specialized tools, while the emerging field of “world models” aims to help AI better understand physical reality—all reflecting a broader shift toward more capable and independent AI systems.
18. Demis Hassabis isn’t shying away from AI’s biggest questions
Fast Company Tech - · May 22
Summary At Google I/O 2026, DeepMind CEO Demis Hassabis presented AI-integrated features across
Google’s product suite while discussing the company’s pursuit of Artificial General Intelligence, which he predicts will arrive by 2030. Hassabis emphasized that while Google is advancing AI capabilities competitively, the company prioritizes addressing AI’s risks and societal impacts, tracing current breakthroughs like agentic AI back to DeepMind’s foundational research. His comments highlight how AI has transitioned from research labs to everyday consumer products, making both its benefits and potential harms increasingly tangible to billions of users.
19. Standard Chartered boss apologises for ‘lower-value human capital’ comments amid job cuts
The Guardian Tech - · May 22
Standard Chartered CEO Bill Winters apologized after facing backlash for describing nearly 8,000 staff facing job cuts due to AI automation as “lower-value human capital.” The bank is cutting about 7,800 back-office roles by 2030 as part of its AI-driven transformation, primarily affecting employees in India, Malaysia, and Poland. Despite Winters’ attempts to clarify and apologize on LinkedIn, many employees and commenters remained critical, viewing his explanations as inadequate justifications for what they considered dehumanizing language.
20. How the AI backlash could cost investors
Axios - · May 22
Growing public backlash against AI—including worker strikes, protests against data centers, and executive booing—poses an underappreciated risk to the technology’s adoption, despite continued massive investor funding. Companies like SpaceX are now formally warning investors in their prospectuses that AI backlash could threaten the sector’s growth trajectory. This matters because widespread resistance could slow AI deployment and create unexpected headwinds for the billions being invested in AI infrastructure and development.
21. Steve Wozniak Tells Graduates They All Have ‘AI’: Actual Intelligence
Slashdot - · May 22
Summary Steve Wozniak delivered a commencement address at Grand Valley State University where he
reassured graduates entering the AI-transformed workforce by reminding them they possess “actual intelligence” — framing AI as humanity’s attempt to replicate brain-like functions through repeated routines. Unlike other speakers who faced backlash for AI hype, Wozniak earned applause by encouraging students to “think different” and avoid following conventional paths in their careers.
22. Abortion Coverage Built on Trust
Columbia Journalism Review - · May 22
Summary Jessica Valenti’s newsletter “Abortion, Every Day” has become an essential resource for
tracking abortion policy and rights across the U.S. after the 2022 Supreme Court overturning of Roe v. Wade, now ranking in Substack’s top ten culture newsletters with a readership of lawyers, activists, providers, and legislators. A recent investigation by Valenti revealed that mothers in multiple states were targeted by Child Protective Services for helping their daughters access abortion, demonstrating the newsletter’s importance as advocacy journalism that prioritizes source protection and accuracy. The story matters because it exposes a critical gap in mainstream media coverage and provides actionable information to those working in reproductive healthcare and rights.
22 stories sourced from Ars Technica, Axios, Columbia Journalism Review, Digital Trends, Fast Company Tech, Futurism, Hacker News, MIT Technology Review, NPR Technology, Slashdot, TechCrunch AI, The Guardian Tech, The Register, The Verge AI. The Slop Report is published daily. Subscribe via RSS.