July 19, 2025

analysis  
Read News
Meta Declines EU AI Code, Google Gemini, Microsoft Favored

Several key developments in the AI landscape have emerged recently. Meta is declining to sign the European Union's AI code of practice, citing legal uncertainties and concerns that it will hinder growth. Joel Kaplan, Meta's Chief Global Affairs Officer, has voiced these concerns, joining over 40 European businesses who have expressed similar sentiments about the EU's AI Act. Other companies like Alphabet and Microsoft have also expressed concerns about the rules, while French AI company Mistral and OpenAI plan to sign the code. The EU has released guidelines to help companies like Google, OpenAI, Meta, Anthropic, and Mistral comply with the AI Act, which will be enforced starting August 2, 2025, for general-purpose AI models and August 2, 2024, for high-risk AI models. These guidelines aim to clarify the rules and reduce burdens for businesses, covering areas like AI model training, provider responsibilities, and transparency. In cybersecurity news, a critical vulnerability called NVIDIAScape has been found in the Nvidia Container Toolkit, affecting 37% of cloud environments and allowing hackers to control AI cloud services. Nvidia has released updates to address this flaw, which has a severity score of 9.0. Elsewhere, Fasikl, a Minnesota startup, has received FDA approval for its Felix NeuroAI Wristband, which uses AI to treat essential tremor. The company has raised over $30 million and plans to commercialize the device. Tech Mahindra plans to invest in AI infrastructure, sales, and marketing over the next 2-3 years, and has opened a 'Manufacturing Xperience Center' in Chennai. Meta has also been hiring AI experts, while Google DeepMind unveiled Gemini Robotics On-Device for real-world AI tasks. However, concerns are rising about AI-generated misinformation, with experts warning that AI is turbocharging the spread of medical misinformation. An attorney in Alabama was removed from a case after using AI to create fake case citations. A portfolio manager suggests Google could lose in the AI race, favoring Microsoft and Broadcom instead. Pennsylvania is set to become an AI hub with over $90 billion in investments, and Nvidia can once again sell semiconductors to China.

Key Takeaways

  • Meta is refusing to sign the EU's AI code of practice due to legal uncertainties and concerns about hindering growth.
  • The EU's AI Act will be enforced starting August 2, 2024, for high-risk AI models and August 2, 2025, for general-purpose AI models, with guidelines released to help companies comply.
  • A critical vulnerability, NVIDIAScape, in the Nvidia Container Toolkit affects 37% of cloud environments, allowing hackers to control AI cloud services; Nvidia has released a fix.
  • Fasikl's Felix NeuroAI Wristband, which uses AI to treat essential tremor, has received FDA approval after raising over $30 million.
  • Tech Mahindra plans to invest in AI infrastructure, sales, and marketing, and has opened a 'Manufacturing Xperience Center' in Chennai.
  • Meta is hiring AI experts, while Google DeepMind has unveiled Gemini Robotics On-Device for real-world AI tasks.
  • Experts warn that AI is making it easier to spread medical misinformation, highlighting the need for regulation.
  • An attorney was removed from a case for using AI to create fake case citations.
  • A portfolio manager suggests Google could lose in the AI race, favoring Microsoft and Broadcom.
  • Pennsylvania is poised to become an AI hub with over $90 billion in investments, and Nvidia can resume selling semiconductors to China.

Meta refuses to sign EU AI code over legal concerns

Meta is refusing to sign the European Union's AI code of practice because of legal uncertainties. Joel Kaplan from Meta says the EU's rules go too far and could hurt AI development in Europe. The EU's code aims to help companies follow the AI Act, which regulates AI. Other tech companies like Alphabet and Microsoft have also expressed concerns about the rules.

Meta rejects EU's AI rules citing legal uncertainties

Meta has decided not to sign the EU's code of practice for general-purpose AI. Joel Kaplan, Meta's chief global affairs officer, says the code has legal uncertainties. The EU's code is meant to help AI companies follow the Artificial Intelligence Act. Companies that don't sign up may face more checks from the European Commission. French AI company Mistral and OpenAI plan to sign the code.

Meta rejects Europe's AI agreement over concerns about growth

Meta will not sign Europe's AI agreement, calling it an overreach that will hurt growth. Joel Kaplan, Meta's global affairs chief, says the code has legal uncertainties. The guidelines are meant to help companies follow the AI Act, which aims to improve transparency and safety. Meta joins other companies in opposing the rules, while Mistral AI has committed to signing the code.

EU gives guidance on AI rules for high-risk AI models

The European Commission has released guidelines to help AI models with systemic risks comply with EU AI rules. These rules affect AI models from companies like Google, OpenAI, Meta, Anthropic, and Mistral. The AI Act will be enforced starting August 2 for high-risk AI models. Companies must evaluate risks, report incidents, and ensure cybersecurity. The guidelines aim to clarify the rules and reduce burdens for businesses.

Meta refuses to sign EU's AI code, calls it overreach

Meta is refusing to sign the European Union's AI code of practice, calling it an overreach. Joel Kaplan, Meta's head of global affairs, says the code has legal uncertainties. The EU's code helps companies follow the AI Act, with copyright protections and transparency rules. Companies that don't sign the code may face more scrutiny. Other European companies also oppose the AI Act.

EU clarifies guidelines for general-purpose AI models

The European Commission has published guidelines for providers of general-purpose AI models. These guidelines clarify the rules for companies using AI in the EU. They cover topics like AI model training, provider responsibilities, and exemptions for open-source AI. The guidelines aim to ensure AI models are transparent and comply with copyright law. The AI Act obligations for general-purpose AI models start on August 2, 2025.

Meta rejects EU's AI code of practice, citing overreach

Meta is refusing to sign the European Commission's AI code of practice, calling it an overreach. Joel Kaplan, Meta's Chief Global Affairs Officer, says the code introduces legal uncertainties. Over 40 European businesses have asked the EU to halt the AI Act. The code aims to help AI companies comply with the AI Act, ensuring AI systems are safe and transparent. OpenAI plans to sign the code.

NVIDIAScape vulnerability threatens AI ecosystem security

Researchers found a critical vulnerability called NVIDIAScape in the Nvidia Container Toolkit. This flaw affects 37 percent of cloud environments, allowing hackers to escape containers. With just three lines of code, attackers can control servers and access sensitive data. This is the second critical vulnerability found in the toolkit within a year. The Nvidia Container Toolkit is used by major cloud providers for AI workloads.

NVIDIA toolkit flaw allows hackers to control AI cloud services

A critical vulnerability in the NVIDIA Container Toolkit, called NVIDIAScape, allows hackers to take over AI cloud services. The flaw, tracked as CVE-2025-23266, has a high severity score of 9.0. It affects 37% of cloud environments, letting attackers steal data and manipulate AI models. A simple three-line exploit can lead to a complete server takeover. NVIDIA has released updates to fix the issue.

AI wristband treats essential tremor, gets FDA approval

Fasikl, a Minnesota startup, created the Felix NeuroAI Wristband to treat essential tremor using AI. The wristband uses AI to adjust electric stimulation and reduce hand tremors. The device recently received FDA clearance after a study with 126 participants. The company has raised over $30 million and plans to commercialize the device. The wristband streams data to the cloud, allowing AI to analyze and adjust stimulation in real time.

Tech Mahindra to invest in AI, sales, and marketing

Tech Mahindra plans to invest in sales, marketing, and AI infrastructure over the next 2-3 years. The company aims to improve its market position and use AI responsibly. Tech Mahindra has opened a 'Manufacturing Xperience Center' in Chennai to help manufacturers use AI. The company also plans to simplify its legal structure and use AI to benefit clients and society.

AI and cybersecurity news: Meta hires, Perplexity valuation, DeepMind robotics

This week saw major developments in AI and cybersecurity. Meta hired more AI experts from Apple for its Superintelligence Labs. Perplexity AI's valuation reached USD 18 billion after new funding. Google DeepMind unveiled Gemini Robotics On-Device for real-world AI tasks. IIT Kanpur's C3iHub launched a cohort to support cybersecurity startups in India.

AI stock bubble may be just beginning, experts say

Experts are discussing whether the AI stock market is in a bubble. Valuations for big tech stocks are high, similar to the dot-com bubble. However, some believe the excitement about AI is justified long-term. They suggest being careful about which companies to invest in. AI is expected to boost productivity across various industries.

Attorney used AI with fake cases, judge removes him

A defense attorney in Loxley, Alabama used an AI program that created fake case citations in a legal brief. The attorney apologized to a federal judge in Mobile. The judge removed the attorney from the case and appointed a new lawyer. The judge also found discrepancies in another motion and will determine a punishment for the attorney.

Google could lose in AI race, says portfolio manager

A portfolio manager suggests that Alphabet's Google could be a loser in the AI race. The manager says they now spend more time using ChatGPT than Google search. They believe Microsoft and Broadcom will benefit greatly from AI. The manager owns stock in both Microsoft and Broadcom.

AI investment in Pennsylvania and Nvidia's China return

This week in tech news includes leaders promising to invest over $90 billion to build an AI hub in Pennsylvania. Also, the Trump administration says Nvidia can sell its semiconductors to China again. Congress is also advancing cryptocurrency bills. Anita Ramaswamy from The Information discussed these details on Marketplace’s “Tech Bytes: Week in Review.”

AI is turbocharging medical misinformation, experts warn

Experts warn that AI is making it easier for medical misinformation to spread. A report from the US Department of Health and Human Services, led by Robert F Kennedy, used AI and contained citation errors. Some researchers believe the report included AI-generated content. AI can create false citations and invent nonexistent material. Experts say there is a need to regulate tech companies to prevent AI misuse.

Sources

Meta European Union AI Act AI code of practice Legal uncertainties Joel Kaplan Alphabet Microsoft Mistral AI OpenAI European Commission AI models Systemic risks High-risk AI models Transparency Safety General-purpose AI models AI model training Provider responsibilities Open-source AI Copyright law NVIDIAScape Nvidia Container Toolkit Vulnerability Cloud environments Cybersecurity Data theft AI models manipulation CVE-2025-23266 Felix NeuroAI Wristband Essential tremor FDA approval AI-powered medical device Tech Mahindra AI infrastructure investment Manufacturing Xperience Center Chennai AI experts Superintelligence Labs Perplexity AI Google DeepMind Gemini Robotics On-Device IIT Kanpur C3iHub Cybersecurity startups AI stock bubble Valuations Dot-com bubble AI investment Productivity AI-generated content Fake case citations Legal brief Google ChatGPT Broadcom Pennsylvania AI hub Nvidia Semiconductors China Medical misinformation