The evening briefing.
Today across AI and tech: AI chatbots struggle with election facts, a memory shortage threatens consumer electronics, and a founding OpenAI member joins Anthropic.
AI Misinformation Risks. A new study reveals AI chatbots, including major rivals to ChatGPT, incorrectly answered election-related questions 90% of the time, raising significant concerns about their use as information sources. This follows issues with Google's AI Overviews, which recently struggled to define basic words, sometimes returning generic chatbot responses instead of definitions. The legal sector also faces challenges, with lawyers continuing to cite fake cases invented by AI, highlighting persistent accuracy problems.
AI Market Dynamics. The AI industry continues to see significant movement, with a founding member of OpenAI joining Anthropic to bolster Claude's research capabilities. This talent shift occurs as both Anthropic and OpenAI focus on enterprise deployments, attracting Wall Street attention. Meanwhile, Qualcomm's stock is surging as investors anticipate a boom in AI devices, signaling a shift towards on-device AI capabilities.
Hardware Supply Chain. A looming memory shortage is expected to cause a significant repricing of consumer electronics in the coming years, driven by fixed wafer capacity among the three major memory manufacturers. However, a recent sighting of Corsair DDR5 sticks using Chinese memory chips offers a potential glimmer of hope that the predicted RAM crisis might end sooner than anticipated. This dynamic highlights the ongoing volatility and strategic importance of the global semiconductor supply chain.
Digital Policy & Security. Regulatory bodies are grappling with the ethical implications of AI, as the US government scrambles to prevent internet users from re-creating dead pilots' voices using AI from crash investigation documents. In a separate security concern, experts discovered that Google API keys can remain usable for over 20 minutes after deletion, creating a significant security vulnerability. These incidents underscore the urgent need for robust policies and improved security practices in the rapidly evolving digital landscape.
AI chatbots got election information wrong 90% of the time in a new study
A new study indicates that popular AI chatbots, including rivals to ChatGPT, made significant errors in election-related information 90% of the time. This raises growing concerns as more individuals rely on AI tools instead of traditional news sources for critical information.
Memory shortage is causing a repricing of consumer electronics
A memory shortage is expected to significantly increase the cost of consumer electronics over the next few years. This is due to the fixed wafer capacity of the three main memory manufacturers, which must be split between different memory types.
Gemini app for Mac adding ‘Spark’ agent and voice control this summer
Google previewed two major features for the Gemini app on macOS at I/O 2026, including a new ‘Spark’ agent and enhanced voice control, which are set to launch this summer.
Meta’s latest app, Forum, looks like Reddit with AI assistants
Meta has launched Forum, a standalone app that combines Facebook Groups with AI assistants, offering a new platform for online discussions and community engagement.
SpaceX will retry Starship test flight after scrubbing prior launch
SpaceX is preparing for another attempt to launch its massive Starship rocket on Friday, following a scrubbed launch plan the previous day.
Blue Origin cleared to fly New Glenn mega-rocket after April mishap
Jeff Bezos' Blue Origin has been cleared to fly its New Glenn mega-rocket after confirming an engine failure led to the loss of an AST SpaceMobile satellite last month.
AI could speed search for brain drugs from decades to years
Researchers are hopeful that AI technology can significantly accelerate the discovery of affordable and effective drugs for conditions like Motor Neurone Disease, potentially reducing development timelines from decades to years.
Qualcomm's stock pop shows investors are 'waking up' to boom in AI devices
Qualcomm's stock is experiencing a significant surge, driven by investor optimism that the company will play a central role in the burgeoning market for AI-powered devices.
Apple shares iPhone and Mac post-quantum cryptography code on GitHub
Apple has published new corecrypto source code on GitHub, along with a detailed technical post explaining its post-quantum cryptography efforts across iPhone, Mac, and other devices.
Google’s AI search is so broken it can ‘disregard’ what you’re looking for
Google's AI Overviews are encountering issues, with searches for terms like "disregard" sometimes yielding generic chatbot responses rather than definitions. This indicates a problem where the AI disregards the user's explicit search query.
US scrambles to stop Internet users re-creating dead pilots’ voices
The US government is working to prevent internet users from re-creating the voices of dead pilots using AI, a practice that circumvents laws against disclosing cockpit audio recordings from crash investigations.
Cloudflare CEO Prince says AI is coming for the measurers
Cloudflare CEO Matthew Prince stated that AI is poised to replace middle management and compliance roles, despite a lack of hard evidence linking recent layoffs to AI rather than efficiency programs.
Experts find Google API keys are still usable, even after you delete them
Security experts have discovered that some Google API keys remain usable for more than 20 minutes after they have been deleted. This creates a significant security vulnerability, as users have no way to quickly revoke access or confirm when the keys stop working.
A founding member of OpenAI has joined Anthropic to boost Claude's research capabilities
Andrej Karpathy, a founding member of OpenAI, has joined Anthropic's AI team to lead pre-training research, aiming to enhance the capabilities of the Claude AI model.