Checklist 427: Hallucinating Current Events
AI Chatbots Under Fire for Spreading Disinformation During LA Protests
During fast-moving protests in LA, social media users shared misleading content—including doctored or unrelated images—while chatbots were increasingly used as fact-checkers. However, they frequently got it wrong:
- National Guard Images: Photos published by the San Francisco Chronicle of National Guard troops sleeping on floors were questioned online. California Governor Gavin Newsom reposted them, only to face backlash and accusations that they were AI-generated or misattributed.
- Grok’s Response: Claimed the photos were from “Afghanistan in 2021” and rejected the Chronicle’s reporting.
- ChatGPT’s Take: Echoed the error, stating one photo was taken during Biden’s Kabul airport evacuation.
This led to further spread of disinformation across platforms like Facebook, Truth Social, and Twitter/X.
Other AI Fumbles
- A repurposed image falsely attributed to civil unrest in Paramount, CA.
- Grok incorrectly insisted it was “a real photo from Paramount, Los Angeles, taken on June 7, 2025,” citing non-existent news sources like ABC7 and Los Angeles Times.
- Even after being corrected (the image was actually from New Jersey), Grok refused to back down, claiming the “evidence strongly supports” its version.
AI’s Fact-Checking Credibility in Crisis
Ken emphasized that this is not about political bias but about reliability. However, the ownership and political affiliations of AI platform owners (e.g., Elon Musk’s dual ownership of Twitter and Grok) were noted as potential factors influencing output.
What’s perhaps most concerning is that people aren’t testing these chatbots—they’re using them as trusted sources. And the chatbots, in turn, confidently deliver falsehoods.
In a throwback to earlier advice on the The Checklist, Ken offered practical ways to fight misinformation:
- Use Real Fact-Checkers such as snopes.com, a top source for debunking internet rumors and disinformation.
- Rely on Multiple Trusted News Sources
- Avoid Sole Reliance on AI for Truth Verification
Chatbots are useful for generating content, drafting emails, or even discussing sourdough starters—though even that last use case proved unreliable.
Apple’s WWDC 2025: AI Gets the Spotlight, While iOS 26 Brings Big Privacy Wins for Kids
Apple Emphasizes On-Device AI, Skims Over Privacy — Except for the Kids
Apple’s 2025 Worldwide Developers Conference (WWDC) largely moved away from its traditional focus on privacy and security, but one area where those themes stood strong was in the realm of child safety and digital protections. According to a detailed report from AppleInsider, iOS 26 will overhaul how Child Accounts are created and controlled—offering greater simplicity and privacy-preserving enhancements for families.
On-Device AI: Apple’s Private Approach
While Apple Intelligence dominated the keynote, the company’s approach to generative AI sets it apart from competitors. Instead of relying on cloud processing, Apple leans heavily into on-device models for AI functionality. Key benefits include:
- Offline Capability: Developer AI tools can function without internet access.
- Local Data Retention: Queries and interactions stay on the device, reducing privacy exposure.
- Developer Integration: Third-party developers will gain access to Apple’s LLMs while maintaining privacy protections.
Though not an ironclad guarantee of security, this marks a strong stance against the data-hungry AI norms seen elsewhere in the tech industry.
Child Protection Front and Center
AppleInsider outlines a slew of new features specifically designed to protect minors across all of Apple’s new “OS 26” family (iOS, macOS, iPadOS, watchOS, visionOS, tvOS).
OS Version Alignment
Apple is aligning all of its platforms to version 26 this fall, streamlining version numbers across devices.
Simplified Child Accounts
- Parents will find it easier to create accurate Child Accounts.
- Under-13 accounts will be automatically linked to Family Sharing.
- Once linked, parental controls activate immediately.
New Age Range Categories
Five distinct age brackets by the end of 2025, to help tailor content and app permissions more appropriately across adolescence.
- 13+
- 16+
- 18+
App-Level Age Privacy
Parents can choose whether to share their child’s age range with apps—never the actual date of birth. These settings can be toggled per app and changed at any time.
More Control Over Communication and Content
App Store Changes
Apps will now clearly label if they include:
- User-generated content
- Messaging features
- Ad tracking
- In-app content controls
Ask to Buy & Age Exceptions
- Parents can override age restrictions with Ask to Buy.
- Crucially, permissions can be revoked at any time, cutting off app access.
Communication Safety Enhancements
- Nudity detection will trigger intervention during FaceTime calls.
- Nude images in Shared Albums will be blurred automatically.
Communication Limits
- Kids can request permission to message new contacts.
- Parents approve or deny requests directly in Messages.
Developer Support: PermissionKit
Apple is introducing a PermissionKit framework to guide developers in mirroring Apple’s own privacy-first permission models. This aims to reduce exploitative app behavior and strengthen platform-wide consistency.
While AI dominated headlines at WWDC, Apple’s continued investment in child safety shows that privacy remains a priority—just not always the one in the spotlight. With iOS 26 and its sibling operating systems, Apple doubles down on trust, especially for its youngest users.
