Anthropic Security Risks Emerge While OpenAI Fixes ChatGPT

The rapid expansion of artificial intelligence brings both significant advancements and pressing challenges, particularly in security and regulation. A recent report from cybersecurity firm Wiz reveals that 65% of leading AI companies, including prominent names like Anthropic and Cohere, are inadvertently exposing sensitive information such as API keys and cloud credentials on GitHub. This oversight creates substantial risks for intellectual property theft and cyberattacks, with many firms struggling to implement timely fixes due to a lack of security disclosure channels. Similarly, the widespread adoption of generative AI tools like ChatGPT and Google's Gemini in businesses is escalating the risk of data leaks and privacy breaches. Executives are increasingly focused on balancing human and AI use, especially after incidents like Samsung's 2023 data leak where employees inadvertently fed sensitive information into public GenAI platforms. Gartner predicts that by 2025, 75% of sensitive data will be exposed due to GenAI use, further compounded by hackers leveraging these tools to create malware. To counter these threats, businesses need robust security, employee education, and clear policies. Regulatory hurdles are also emerging, as seen with Google Photos' new Conversational Editing AI tool, which is unavailable to users in Texas and Illinois. This restriction likely stems from state biometric data laws, specifically Illinois' BIPA and Texas' Capture or Use of Biometric Identifier Act, which regulate the collection of facial geometry data by features like Google's Face Groups. Meanwhile, AI continues to evolve on the user experience front, with OpenAI addressing a common user complaint by allowing ChatGPT users to remove em dashes from generated text, a fix confirmed by CEO Sam Altman. On the infrastructure side, the demand for AI data centers is soaring, with global spending projected to hit $580 billion this year, surpassing investments in new oil supplies. This surge places immense pressure on existing electrical grids and highlights a critical need for sustainable power solutions, with half of this new electricity demand expected from the US. Companies like OpenAI are planning massive investments in chips and energy to support their ambitious AI initiatives. However, these developments are not without local opposition; residents in Saline Township, Washtenaw County, Michigan, are strongly resisting a proposed multi-billion-dollar AI data center, citing concerns over water usage and environmental impact, despite the developer, Related Digital, agreeing to protections and investing $14 million in local initiatives. Beyond infrastructure, AI is transforming industries and corporate strategies. Google's DeepMind AI is already making a tangible impact by improving hurricane forecasting speed and accuracy, proving its capability by predicting Hurricane Melissa's rapid intensification and Category 5 landfall in Jamaica. While meteorologists are impressed, some hope Google will offer more insight into the model's workings. In the corporate world, Meta is integrating AI deeply into its operations, announcing that starting in 2025, employee performance will be linked to 'AI-driven impact' to accelerate its transition to an 'AI-native' company. Employees are encouraged to highlight 'AI-fuelled wins' in their self-evaluations, with tools like Meta's AI Performance Assistant and suggestions to use Metamate or Google's Gemini for content. Looking ahead, Shanghai is planning to revolutionize its dining sector by creating 'AI restaurants' with automated kitchens and robot servers by 2028, aiming for over 70% of fast-food and drink chains and over 50% of full-service restaurants to adopt smart technologies. This initiative seeks to establish Shanghai as a world-class hub, though experts ponder the potential loss of human touch and customer enjoyment.

Key Takeaways

• 65% of leading AI companies, including Anthropic and Cohere, are accidentally leaking sensitive information like API keys and cloud credentials on GitHub.
• Generative AI tools like ChatGPT and Google's Gemini are increasing business data leak risks, with Gartner predicting 75% of sensitive data exposure by 2025 due to GenAI use.
• Google Photos' AI editing tool, Conversational Editing, is unavailable in Texas and Illinois due to state biometric data laws related to its Face Groups feature.
• OpenAI has updated ChatGPT to allow users to remove em dashes from generated text, addressing a common user complaint.
• Google's DeepMind AI is significantly improving hurricane forecasting speed and accuracy, successfully predicting Hurricane Melissa's rapid intensification.
• Meta will link employee performance to 'AI-driven impact' starting in 2025 to accelerate its goal of becoming an 'AI-native' company.
• A proposed multi-billion-dollar AI data center in Washtenaw County, Michigan, faces strong local opposition over environmental concerns, despite developer efforts.
• Global spending on data centers, driven by AI, will reach $580 billion this year, surpassing new oil supplies and creating immense demand for renewable energy.
• OpenAI plans massive investments in chips and energy to power its expanding AI initiatives.
• Shanghai plans to establish 'AI restaurants' with automated kitchens and robot servers by 2028, aiming to become a world-class hub for smart dining technologies.

Top AI Companies Accidentally Leak Secrets on GitHub

Cybersecurity firm Wiz found that 65% of leading AI companies are accidentally leaking sensitive information on GitHub. These leaks include API keys, authentication tokens, and cloud credentials. Companies like Perplexity, Anthropic, and Cohere are among those implicated. This exposure could allow unauthorized access to their AI models and user data, risking intellectual property and cyberattacks. Many firms were slow to fix these issues due to a lack of security disclosure channels and limited resources. This problem highlights a critical security oversight in the fast-paced AI industry.

Generative AI Tools Increase Business Data Leak Risks

Generative AI tools like ChatGPT and Gemini are becoming common in businesses but also increase risks of data leaks and privacy breaches. A report notes 71% of executives now focus on balancing human and AI use to reduce these threats. Employees accidentally leak sensitive information by inputting it into public GenAI platforms, as seen with Samsung in 2023. Gartner predicts that by 2025, 75% of sensitive data will be exposed due to GenAI use. Hackers also use GenAI to create malware, adding to cyber threats. Businesses must use strong security, educate employees, and implement clear policies to protect their data.

Google Photos AI Editing Blocked in Texas and Illinois

Google Photos' new AI editing tool, Conversational Editing, is not available to users in Texas and Illinois. This restriction likely stems from Google's Face Groups feature, which collects biometric data like facial geometry. Both states have laws, Illinois' BIPA and Texas' Capture or Use of Biometric Identifier Act, that strictly regulate the collection and storage of such data. These laws typically require explicit consent, and violations can lead to significant penalties. Lawyers explain that the issue is not the editing tool itself, but the underlying biometric data collection. Google has not commented on the situation.

OpenAI Lets ChatGPT Users Remove Em Dashes

OpenAI has updated ChatGPT, allowing users to remove em dashes from generated text. This change addresses a "most hated writing habit" of the AI tool. Sam Altman, OpenAI's CEO, confirmed the fix. Users now have more control over the punctuation in their ChatGPT outputs. This small but important update improves the user experience for many.

Washtenaw County Residents Oppose AI Data Center Plan

Residents in Saline Township, Washtenaw County, Michigan, are upset about a proposed multi-billion-dollar AI data center. The township board initially voted against rezoning but faced lawsuits from the developer, Related Digital, and landowners. They then agreed to a consent judgment that includes protections for noise, traffic, and water usage, like a closed-loop water system. Related Digital also plans to invest $14 million in local initiatives. Despite the developer's efforts to engage the community, many residents feel blindsided and distrust the project's impact, especially on water resources. DTE Energy is pushing for a final decision by December 5, with construction planned for early next year.

Meta Links Employee Performance to AI Impact in 2025

Meta announced it will connect employee performance to "AI-driven impact" starting next year. Janelle Gale, Meta's head of people, stated this move aims to recognize those who help the company become "AI-native" faster. For 2025, employees will be rewarded for exceptional AI contributions in their work or team improvements. While AI metrics will not be formal in 2025 reviews, staff should include "AI-fuelled wins" in their self-evaluations. Meta also provides an AI Performance Assistant for managers and suggests using Metamate or Google's Gemini for content.

Google DeepMind AI Improves Hurricane Forecasting Speed

Google's DeepMind AI tool is making hurricane forecasting faster and more accurate, potentially saving lives. National Hurricane Center meteorologists, like Philippe Papin, are using the DeepMind model, first released in June. This machine learning tool quickly spots weather patterns that traditional physics-based models might miss. It proved its accuracy by predicting Hurricane Melissa's rapid intensification and Category 5 landfall in Jamaica. Experts are impressed with DeepMind's performance, noting it is less expensive and time-consuming to run. However, some forecasters hope Google will provide more insight into how the "black box" model reaches its predictions.

AI Data Centers Boost Renewable Energy Demand

A new report from the International Energy Agency shows that $580 billion will be spent on data centers this year, surpassing spending on new oil supplies. These AI data centers will require immense power, putting stress on existing electrical grids. However, this boom could also greatly benefit companies focused on renewable energy and sustainable data center design. Half of this new electricity demand will come from the US, with many centers located near large cities. Companies like OpenAI plan massive investments in chips and energy, highlighting the need for both corporate and government efforts to power these facilities sustainably.

Shanghai Plans AI Powered Restaurants and Robot Kitchens

Shanghai has unveiled a plan to create "AI restaurants" featuring automated kitchens and robot servers, aiming to become a world-class hub by 2028. The city wants over 70% of fast-food and drink chains, and over 50% of full-service restaurants, to use smart technologies. This shift could help large restaurant chains but might challenge smaller, family-run eateries. The plan also suggests moving towards fewer staff in some operations. However, experts worry that diners might not like the loss of human touch or a "machine taste," which could reduce customer enjoyment. Shanghai hopes this initiative will create a new dining ecosystem that other cities might adopt.

Sources

• AI Titans’ GitHub Gaffes: Leaking Secrets in the Race to Innovate
• Generative AI’s Silent Sabotage: Data Leaks Threaten Business Empires
• A Google Photos AI Editing Feature Isn't Available in These States. We Asked Lawyers Why
• OpenAI ends ChatGPT’s most hated writing habit, allowing users to remove em dashes from text
• Controversy grows over AI data center project in Washtenaw County
• Meta To Tie Employee Performance To “AI-Driven Impact” Starting Next Year
• How Google’s DeepMind tool is ‘more quickly’ forecasting hurricane behavior
• How much of the AI data center boom will be powered by renewable energy?
• Shanghai dreams of a new era of ‘AI restaurants’. But can diners stomach it?