OpenAI ChatGPT hallucinates product picks as Donald Trump halts regulation

Key Takeaways

• President Donald Trump's administration is actively discouraging state-level AI regulation, threatening to withhold federal broadband funding from states like Louisiana, which subsequently scaled back several proposed AI bills.
• The White House intervened to stop AI regulatory bills in Utah, advocating for a unified national approach to AI policy rather than state-specific laws.
• OpenAI's ChatGPT has been found to frequently hallucinate product recommendations, incorrectly attributing them to WIRED reviews for items like TVs and headphones, raising concerns about AI reliability for purchase decisions.
• Wikipedia's English-language editors have banned all AI-generated text from articles to preserve content integrity, though AI tools can still be used for proofreading or translation.
• San Jacinto College launched a new AI certification program for individuals with an associate degree or higher, aiming to make AI fundamentals more accessible to professionals.
• Evotec appointed Ashiq Khan, an expert in robotics and AI and co-founder of Iktos, as its chief commercial officer to enhance its commercial execution and growth in biopharmaceutical services.
• Soxton AI acquired Cipher, a company specializing in real-time security for autonomous AI applications, to strengthen its legal technology platform and protect sensitive startup data.
• Intuit's 'AI-HI' approach, combining AI agents with human expertise, achieved an 85% repeat usage rate, demonstrating the effectiveness of hybrid models for customer confidence and task automation.
• Organizations are experiencing unexpectedly high operational costs for AI-native hardware like GPUs and TPUs, alongside significant challenges in thermal management and power consumption.
• New studies are exploring solutions for 'AI Brain Fry,' a reported mental exhaustion from extensive AI tool usage, while patent experts advise long-term strategies for drafting AI patents to navigate evolving regulations.

Louisiana scales back AI rules after Trump threats

Louisiana lawmakers have withdrawn some, but not all, proposed regulations for artificial intelligence (AI). This action follows threats from President Donald Trump to withhold federal broadband funding from states that regulate AI. At least seven bills related to AI have been shelved by legislators. Some lawmakers are still pursuing other AI-related bills, believing regulation is eventually necessary. The withdrawn bills covered areas like AI in health insurance, identity recreation, and employment decisions. Bills that may still advance include those requiring consent for AI transcription of medical conversations and disclosure of AI use in automated calls.

White House stops Utah AI bill, seeking no AI regulation

The White House has intervened to stop artificial intelligence (AI) regulatory bills proposed by Utah Republican Representative Doug Fiefia. This action came after criticism from the Trump administration and industry leaders who favor minimal regulation. The administration's stance suggests a preference for a unified, national approach to AI policy rather than state-specific laws. The intervention highlights the ongoing debate about how AI technologies should be governed and the potential impact of federal policy on state-level initiatives.

Louisiana AI proposals reduced after Trump funding threat

Louisiana has withdrawn several proposed regulations for artificial intelligence (AI) following threats from President Donald Trump to cut federal broadband funding. While many bills are being shelved, some legislators plan to continue with their AI proposals. There is confusion about which AI measures might conflict with Trump's executive order. Bills that have been stopped include those limiting AI in health insurance and medical treatment. Some lawmakers are still pushing for AI regulations, believing they are eventually needed.

ChatGPT incorrectly recommends products from WIRED reviews

An investigation found that ChatGPT frequently provides incorrect product recommendations, even when asked about specific reviews from WIRED. The AI chatbot hallucinated product picks for TVs, headphones, and laptops that WIRED reviewers did not recommend. This inaccuracy raises concerns about the reliability of AI for purchase decisions. OpenAI stated they are improving product discovery tools, but visiting the source website remains the most dependable method for accurate information. The findings highlight the ongoing issue of AI hallucination, where chatbots generate plausible but false information.

ChatGPT fails product tests, invents WIRED recommendations

A recent investigation revealed that ChatGPT fabricates product recommendations, presenting them as WIRED's top picks for TVs, headphones, and laptops. These recommendations were not endorsed by WIRED's reviewers, indicating a significant issue with AI accuracy. Consumers relying on ChatGPT for buying advice could be misled by this 'hallucination' of information. The study emphasizes that large language models generate text based on patterns, not factual retrieval, leading to potential inaccuracies. Users are advised to verify AI-generated recommendations with original sources due to these ongoing reliability problems.

San Jacinto College offers new AI certification

San Jacinto College has launched a new certification program focused on artificial intelligence (AI). This program is designed for individuals who already possess an associate degree or higher. The coursework will cover AI fundamentals and its various applications. The college aims to demystify AI and make it more accessible to professionals. This initiative prepares individuals for the growing demand for AI expertise in the workforce.

New studies offer solutions for 'AI Brain Fry'

Two new studies suggest potential ways to help individuals experiencing 'AI Brain Fry,' a feeling of mental exhaustion from using AI tools. Researchers observed that while AI tools increase productivity, users also reported reaching their cognitive limits. The studies aim to provide solutions for this growing concern as more people integrate AI into their work. The findings offer reassurance that methods exist to manage the mental strain associated with extensive AI use.

Evotec hires AI expert for commercial team lead

Evotec has appointed Ashiq Khan, Ph.D., as its new chief commercial officer, placing an executive with robotics and artificial intelligence (AI) experience at the head of its commercial unit. This move is part of Evotec's recently announced transformation plan, which emphasizes commercial execution. Khan's background in driving growth in AI and drug discovery platforms is expected to help Evotec identify and create new opportunities. He previously co-founded Iktos, an AI-driven drug discovery company. This appointment aims to strengthen Evotec's position in the biopharmaceutical services market.

Wikipedia bans AI-generated text from articles

Wikipedia's English-language editors have voted to ban all AI-generated text from its articles, effective immediately. While AI tools can still be used for proofreading or translation, original content must be human-written. This decision addresses concerns about the proliferation of AI-generated content, including fabricated citations and unnatural language. The policy aims to maintain the integrity and reliability of Wikipedia's information. Editors will focus on updating guidelines and addressing AI agent policies.

Soxton AI buys Cipher to enhance security

Soxton AI has acquired Cipher, a company specializing in real-time security for autonomous AI applications. This acquisition strengthens Soxton AI's legal technology platform by integrating Cipher's security measures to protect sensitive startup data. Cipher's team will join Soxton, boosting their capabilities in building multi-agent systems for legal processes. The acquisition was completed using Soxton AI's own platform, demonstrating its ability to automate legal workflows and reduce costs. This move enhances the safety and reliability of AI-powered legal systems for startups.

Draft AI patents to last through changing rules

Drafting artificial intelligence (AI) patents requires a long-term strategy that anticipates future guidance changes from patent offices. Experts advise focusing on the core invention story and expanding the patent specification to create fallback positions. Instead of broad initial claims, it's recommended to file narrow, defensible claims first, with broader claims following in continuations. Careful word choice is crucial to avoid triggering rejections based on mental steps. The goal is to create patents that remain valid and enforceable despite evolving patentability standards for AI technologies.

AI hardware costs are increasing unexpectedly

Organizations investing in artificial intelligence (AI) are facing unexpectedly high operational costs, particularly with AI-native hardware. These specialized systems, like GPUs and TPUs, are efficient for AI tasks but can be inefficient for general computing. Challenges include operational ambiguity, where it's unclear if issues stem from infrastructure, data, or the AI model itself. Thermal management and power consumption are also significant concerns, often exceeding initial forecasts. Data center operators must carefully balance AI hardware with general-purpose servers to manage workloads and costs effectively.

Intuit AI agents succeed by including humans

Intuit's AI agents have achieved an 85% repeat usage rate by combining artificial intelligence with human expertise. This 'AI-HI' approach provides customers with greater confidence and trust in the financial software. The AI agents assist with tasks like closing books, running payroll, and identifying discrepancies, leading to faster invoice payments and reduced manual work. Intuit ensures human experts are always accessible for complex issues or when users need to discuss ideas. This hybrid model proves that human involvement remains crucial for effective AI implementation in enterprise environments.

Sources

• Louisiana scraps some, but not all, AI proposals after Trump threats
• ‘They want zero regulation’: White House trumps Utah Republican’s bill to regulate AI
• Louisiana scraps some, but not all, AI proposals after Trump threats
• I Asked ChatGPT What WIRED’s Reviewers Recommend—Its Answers Were All Wrong
• ChatGPT Fails Product Recommendation Test, Hallucinates WIRED Picks
• San Jacinto College launches artificial intelligence certification
• Worried You Might Have 'AI Brain Fry?' 2 New Studies Suggest We Might Know How to Help
• Evotec hires exec with AI experience to lead rebooted commercial team
• One of the Internet’s Most Iconic Websites Just Took a Bold Stand. The Rest Should Follow.
• Soxton AI Acquires Cipher To Add Agentic Security
• How to Draft AI Patents That Survive the Next Guidance Cycle, and the One After That
• How AI costs are quietly spiraling
• Intuit's AI agents hit 85% repeat usage. The secret was keeping humans involved