Google Chrome quietly downloads 4GB AI model Gemini Nano without user permission

Key Takeaways

• Google Chrome is quietly downloading a 4GB AI model called Gemini Nano onto users' devices without permission, potentially violating EU laws.
• Over 35% of websites created since 2022 were generated by AI, making the internet less diverse in language and more positive overall.
• Colorado lawmakers propose HB26-1263 to regulate conversational AI with flexibility for rapid tech changes.
• Illinois lawmakers debate two AI liability bills: one backed by OpenAI limits developer liability, another backed by Anthropic requires independent safety audits.
• Houston ISD will open nine AI-focused schools in 2024, teaching core subjects in the morning and hands-on skills in the afternoon.
• Arctic Wolf laid off 250 workers (less than 10% of workforce) to invest more in its AI superintelligence platform.
• Climate TRACE database may underestimate urban vehicle CO2 emissions by an average of 70.4%, according to a study; co-founder Gavin McCormick disputes the findings.
• HHS official Dr. Thomas Keane says AI can help doctors clinically, and a drug pricing plan is expected to save $529 billion.
• Leaders should match AI tools to decision types: analytical AI for narrow problems, generative AI for wide problems.
• AI companies should stop using human-like names for features (e.g., dreaming, memories) to avoid misleading users.

Google Chrome downloads 4GB AI model without asking users

Security researcher Alexander Hanff claims Google Chrome is quietly downloading a 4GB AI model called Gemini Nano onto users' devices without their permission. The download happens in the background while the browser is idle and does not ask for user consent. Hanff says this practice may violate EU laws like the ePrivacy Directive and GDPR. Google says the model powers security features like scam detection and can be turned off in settings, but critics argue users should have to opt in first.

Google Chrome stores 4GB AI model on devices without asking

Security researcher Alexander Hanff reports that Google Chrome is quietly storing a 4GB AI model called Gemini Nano on devices that meet minimum hardware requirements. The browser does not ask for permission to install the model, and there is no easy way to opt out. If users delete the files manually, Chrome will redownload them automatically. Google says the model powers security features and can be turned off in settings, but critics argue users should have a choice.

AI creates 35 percent of new websites since 2022

Researchers from Stanford, Imperial College London, and the Internet Archive found that over a third of websites created since 2022 were generated by AI. They tested six common criticisms of AI text and found only two were true: AI makes the internet less diverse in language and more positive overall. The study did not find that AI increases lies or fails to cite sources. The author argues that AI has no interest in truth, justice, beauty, or love, and urges people to label AI-generated content clearly.

Colorado lawmakers push for AI regulation that adapts quickly

Colorado lawmakers are working on AI regulation that can keep up with the fast pace of technology. They argue that waiting for perfect answers allows harm to happen, as seen with social media. The proposed bill HB26-1263 aims to set clear rules for conversational AI while leaving room for changes as the technology evolves. The goal is to create a framework that protects consumers, especially youth, without waiting for a perfect solution.

Houston ISD opens nine AI-focused schools next year

Houston ISD will launch nine AI-focused campuses next year, including five elementary and four middle schools. The schools will teach core subjects in the morning and focus on hands-on skills in the afternoon. The district says this prepares students for a future where AI is common and some jobs may be replaced. Parents who do not want their children in AI-focused schools can choose a different campus.

Arctic Wolf cuts 250 jobs to invest more in AI

Cybersecurity company Arctic Wolf laid off 250 workers, which is less than 10 percent of its workforce. The company says the restructuring will help it invest more in its AI superintelligence platform and agentic Security Operations Center. The cuts affected sales, product development, and marketing roles. Arctic Wolf competes with companies like CrowdStrike and SentinelOne in the cybersecurity market.

Study says AI carbon database may underestimate vehicle emissions

A new study claims that the Climate TRACE database, which uses AI to track carbon emissions, may drastically underestimate CO2 from vehicles. Researchers from Northern Arizona University compared Climate TRACE data to the Vulcan Project and found an average difference of 70.4 percent lower for urban vehicle emissions. Climate TRACE co-founder Gavin McCormick disagrees and says their data matches official U.S. inventories. The study urges policymakers to use the data with caution.

HHS official says AI can help doctors in clinical settings

Dr. Thomas Keane, the National Coordinator for Health Information Technology, said AI is moving into areas where it can help doctors clinically. He discussed a drug pricing plan expected to save $529 billion and the use of AI in medicine on Fox Business. The video clip highlights the growing role of AI in healthcare.

Illinois lawmakers debate AI liability for catastrophes

Illinois lawmakers are considering two competing bills about who is responsible when AI causes serious harm or death. One bill, supported by OpenAI, would limit developer liability if they follow safety plans. Another bill, supported by Anthropic, would require independent safety audits and reporting of serious incidents. Both proposals face a May 15 deadline for a vote in the Illinois Senate.

Match AI tools to the type of decision you need to make

An article from MIT Sloan Management Review argues that leaders should use different AI tools for different types of decisions. Analytical AI is good for narrow problems with clear data and goals, like where to open new stores. Generative AI is better for wide problems that need exploration and alignment, like a brand pivot. Using the wrong type of AI can lead to poor results and weak buy-in.

Author begs AI companies to stop using human names for features

A Wired article criticizes AI companies for naming features after human processes like dreaming, memories, and imagination. The author argues that this creates a false idea of what AI can do and is misleading. AI is a tool, not a human, and should be described honestly. The article calls for clearer and more accurate language about AI capabilities.

Sources

• Google Chrome 'silently' downloads 4GB AI model to your device without permission, report claims — researcher says practice may violate EU law, waste thousands of kilowatts of energy
• Google Chrome Is Downloading a 4GB AI Model Onto Your Device Without Consent, Researcher Warns
• Artificial Intelligence Creates 35% of New Websites, Does Not Seek Truth, Justice, Beauty, or Love
• Legislating at the speed of artificial intelligence
• Houston ISD will launch nine AI-focused campuses next year. Here’s what we know.
• Arctic Wolf cuts 250 jobs in AI push
• This AI-Powered Carbon Emissions Database Has an Embarrassing Error, Study Claims
• We are moving into areas where AI can help doctors clinically, HHS official says
• Illinois lawmakers debating competing proposals on AI liability in catastrophes
• Calibrate AI Use to the Decision at Hand | MIT Sloan Management Review
• I Am Begging AI Companies to Stop Naming Features After Human Processes