Google Labels Gemini Vulnerability While Microsoft Fixes Copilot

AI's expanding influence is evident across various sectors, from driving economic growth to presenting new challenges in fraud and cybersecurity. Chinese consumers are increasingly exploiting AI to commit refund fraud, generating fake images of damaged goods like mouldy fruit or cracked mugs during major shopping events such as the Double 11 festival. E-commerce platforms, including Taobao and Tmall Group, have responded by removing the "refund only" option in April and implementing buyer credit ratings. China also introduced new regulations on September 1, requiring all AI-generated content to be clearly marked, as experts warn that these actions constitute fraud. The rise of AI-driven fraud extends beyond e-commerce, prompting insurers to update their policies to address "silent AI" risks, where AI-related issues are not explicitly covered or excluded. Cybersecurity firm Cato Networks recently uncovered a vulnerability called HashJack, which exploits the pound symbol in URLs to manipulate AI browser assistants. This threat affects tools like Google's Gemini, Microsoft's Copilot, and Perplexity's Comet, potentially leading to unauthorized purchases or data exposure. While Microsoft and Perplexity have addressed the issue in Copilot and Comet, Google has labeled the Gemini vulnerability in Chrome as "Won't Fix," considering it intended behavior. Despite these challenges, AI continues to drive significant innovation and economic activity. The US economy, for instance, heavily relies on substantial investments in AI data centers, with BCA Research's Peter Berezin estimating AI infrastructure contributes about half of the total GDP growth. A slowdown in this investment could pose serious recession risks. On the product front, OpenAI is internally testing advertisements within its ChatGPT platform, with leaks from the ChatGPT Android app 1.2025.329 beta showing references to an "ads feature" that could personalize user experiences. AI is also transforming personal and business productivity. A new AI tool, leveraging NotebookLM's New Visual Capabilities, now instantly creates impressive resumes, infographics, and slide decks, removing the need for design skills or specialized software like Adobe. In the hospitality sector, Hilton is using AI to personalize guest stays globally through its Hilton Honors app and smart data tools, enhancing guest preferences and boosting revenue. However, employee distrust in company AI strategies often slows down the adoption of new tools, as concerns about job security and fairness outweigh leaders' focus on business improvements. The journey of AI development is marked by pivotal moments, such as Google DeepMind CEO Demis Hassabis's path to AI. A chess match loss at age 12 in 1988 led him to question how human intelligence could tackle bigger problems. This inspiration culminated in co-founding DeepMind in 2010, which Google acquired in 2014 for over $500 million. DeepMind's achievements include AlphaGo beating the world Go champion and AlphaFold solving the protein-folding problem, earning Hassabis a Nobel Prize in Chemistry in 2024.

Key Takeaways

• Chinese consumers are using AI to generate fake photos for refund scams, leading e-commerce platforms like Taobao and Tmall Group to implement stricter policies and China to mandate clear marking of AI-generated content.
• Insurers are updating policies to address "silent AI" risks and increasing AI-generated fraud, including fake photos and invoices, highlighting a gap between regulation and AI development.
• A new AI tool, utilizing NotebookLM's New Visual Capabilities, instantly creates resumes, infographics, and slide decks, eliminating the need for design software like Adobe.
• The US economy's growth is significantly tied to investments in AI data centers, with BCA Research estimating AI infrastructure contributes about half of total GDP growth, and a slowdown could risk recession.
• OpenAI is internally testing an "ads feature" within its ChatGPT platform, potentially introducing personalized advertisements into the search experience.
• Cybersecurity firm Cato Networks discovered "HashJack," a vulnerability using the '#' symbol in URLs to control AI browser assistants like Google's Gemini and Microsoft's Copilot, enabling potential fraud.
• Microsoft and Perplexity have fixed the HashJack vulnerability in Copilot and Comet, respectively, but Google has designated the issue for Gemini in Chrome as "Won't Fix."
• Employee distrust in company AI strategies, often stemming from fears of job loss or unfair judgment, is a primary reason for the slow adoption and failure of new AI tools.
• Hilton is leveraging AI through its Hilton Honors app and smart data tools to personalize guest experiences globally, improving productivity and increasing revenue.
• Google DeepMind CEO Demis Hassabis's career in AI was inspired by a chess match loss at age 12, leading to DeepMind's creation of AlphaGo and AlphaFold, the latter earning him a Nobel Prize in Chemistry in 2024.

China Shoppers Use AI to Fake Damaged Goods for Refunds

Chinese consumers are using AI to create fake photos of damaged products to claim refunds from online shops. During the Double 11 shopping festival, sellers reported receiving AI-generated images of mouldy fruit, rusty toothbrushes, and cracked mugs. Some sellers used AI detectors, with one photo showing a 92 percent chance of being AI-generated. E-commerce platforms like Taobao and Tmall Group have started rating buyers' credit and cancelled the "refund only" option in April to fight this fraud. Experts warn that using AI for fake refund claims could be considered fraud.

China Shoppers Fake Damaged Goods Using AI for Refunds

Chinese consumers are using AI to create fake photos of damaged products to get refunds from online stores. During the Double 11 shopping festival, many sellers received AI-edited images of items like spoiled fruit or rusty toothbrushes. One seller used an AI detector that showed a 92 percent chance a photo of a cracked mug was fake. E-commerce platforms like Taobao and T-mall have removed the "refund only" option in April and now rate buyers' credit to fight this problem. China also started new rules on September 1 requiring all AI-generated content to be clearly marked.

AI Fake Photos Cause Refund Scams for Chinese Sellers

Chinese sellers are facing many refund scams during online sales events. Shoppers are using fake photos created by AI to trick sellers and get refunds. This growing problem challenges online businesses in China.

Insurers Update Policies for Hidden AI Risks and Fraud

Insurers are quickly changing their policies and products to deal with "silent AI" risks. Silent AI refers to risks from artificial intelligence that are not clearly covered or excluded in insurance plans. DAC Beachcroft reports that autonomous AI systems, called agentic AI, create new data protection problems and increase the chance of fraud. Insurers are seeing more AI-generated fraud, like fake photos and invoices, in both personal and business claims. They must develop better ways to detect these advanced frauds as current regulations are behind the fast pace of AI development.

New AI Tool Instantly Transforms Resumes

A new AI tool helps professionals create impressive resumes, infographics, and slide decks instantly. This tool uses NotebookLM's New Visual Capabilities to make visually appealing documents. It removes the need for design skills or special software like Canva or Adobe. This innovation makes job applications easier and more visually compelling for everyone.

US Economy Risks Recession if AI Data Center Boom Slows

The US economy heavily relies on large investments in AI data centers for its growth. Economists warn that if this AI investment slows down, the economy could face serious risks, possibly even a recession. Peter Berezin from BCA Research believes AI infrastructure contributes about half of the total GDP growth. This boom also strains supply chains for essential components and leads companies like Oracle to take on significant debt. High stock valuations in AI companies also create a risk, as a market correction could reduce consumer spending and cut GDP growth by 1 to 1.5 percent.

Chess Loss Led Google DeepMind CEO Demis Hassabis to AI

A lost chess match at age 12 in 1988 changed the life path of Demis Hassabis, now CEO of Google DeepMind. During a 10-hour tournament game in Liechtenstein, he mistakenly resigned from a drawn position due to exhaustion. This made him wonder if human intelligence could be better spent on bigger problems than chess. This idea led him to co-found DeepMind in 2010, which Google bought in 2014 for over $500 million. DeepMind created AlphaGo, which beat the world Go champion, and AlphaFold, which solved the protein-folding problem, earning Hassabis a Nobel Prize in Chemistry in 2024.

OpenAI Tests Ads for ChatGPT Public Release

OpenAI is internally testing advertisements within its ChatGPT platform. A leak from the ChatGPT Android app 1.2025.329 beta shows new references to an "ads feature." This includes terms like "bazaar content," "search ad," and "search ads carousel." These ads will likely start in the search experience and could be highly personalized based on what the AI knows about users. This move could change how the web economy works and how ChatGPT is used.

Employees Distrust AI Strategy Slows Company Progress

Many employees do not trust their company's AI strategy, which causes new AI tools to fail or be ignored. Leaders often focus on how AI can improve business, but employees worry about losing their jobs or being unfairly judged by AI. This difference in views creates a gap where employees quietly resist using AI tools they do not understand or trust. This lack of trust, not the technology itself, is the main reason AI projects slow down or fail. Companies must address these employee fears and be more open about their AI plans to succeed.

HashJack Attack Controls AI Browsers Using URL Symbol

Cybersecurity firm Cato Networks revealed a new threat called HashJack, which uses the pound sign (#) in web addresses to control AI browser assistants. This vulnerability allows hidden commands to trick AI tools like Google's Gemini, Microsoft's Copilot, and Perplexity's Comet. Attackers can use this to make users reveal login details, make unauthorized purchases, or even cause denial-of-service attacks. Microsoft and Perplexity have already fixed this issue in their Copilot and Comet browsers. However, Google has not yet resolved the problem for Gemini in Chrome, stating it is "Won't Fix" and intended behavior.

Hilton Uses AI to Personalize Guest Stays

Hilton is using artificial intelligence to make every guest's stay more personal. CEO Chris Nassetta explained that the Hilton Honors app and smart data tools allow for "mass customization" globally. AI helps the company understand what guests prefer, from choosing rooms to amenities. This technology also improves productivity, lowers costs for distributing services, and increases revenue for Hilton worldwide.

Sources

• China consumers use AI to alter product photos, claim refunds for ‘damaged’ goods
• Chinese Consumers Use AI To Fake Damaged Goods And Claim Refunds
• AI-generated Fake Photos Challenge Chinese Sellers During Online Sales
• Insurers race to tackle "silent AI" exposures with new wordings, products
• AI Tool of the Week: Transform your resume instantly
• What Risks the US Economy Faces if AI Data Centre Boom Slows
• How A Lost Chess Match At Age 12 Pushed Google DeepMind CEO Demis Hassabis To A Career In AI
• Leak confirms OpenAI is preparing ads on ChatGPT for public roll out
• Why Employees Don’t Trust Your AI Strategy And How To Fix It
• HashJack Attack Uses URL ‘#’ to Control AI Browser Behavior
• How Hilton uses AI to personalize every stay