Consumers Prefer Human-Designed Products Over AI, Warns Hinton

Key Takeaways

• Georgia State University research shows consumers rate AI-designed products as less sustainable than human-designed ones.
• More than three quarters of people chose a human-designed toothbrush over an AI-designed equivalent in testing.
• Geoffrey Hinton estimates a 10 percent chance AI could kill all humans within a decade.
• Lawmakers will hold a hearing on AI existential risks on September 11, following a statement from over 1,000 experts.
• Proofpoint will add Microsoft 365 connectivity to its AI security tools in the fourth quarter of 2026.
• A report from Anthropic details how hackers used Claude to steal drone software and gain cloud control in three hours.
• Zettabyte and Raku Advanced Tech signed a deal on September 11, 2026, to build green AI data centers in Taiwan and Thailand.
• A Mysterium VPN study found 36,769 self-hosted AI systems exposed to the public internet without authentication.
• New EU rules starting in February 2025 require companies to tailor AI training to specific job roles.
• Security firms like Infinigate and KnowBe4 are expanding partnerships to cover AI agent protection in European markets.

Study finds consumers doubt AI designed products

A new study from Georgia State University shows that people view products designed by AI as less sustainable than those made by humans. In tests, an identical backpack scored lower on sustainability when labeled as AI designed, and more than three quarters of people chose a human designed toothbrush over an AI one. Researchers say this happens because consumers link sustainability to human care and moral intent, which they do not assume machines possess. The study suggests companies should explicitly communicate the values behind their AI design choices to reduce this skepticism.

Study shows consumers doubt AI designed products

Research from Georgia State University and the University of South Florida reveals that consumers rate products as less sustainable when they know AI designed them. Experiments showed that identical items received lower sustainability scores and fewer clicks when marketed as AI designed compared to human designed versions. The study identifies a lack of perceived genuine care as the main reason for this distrust among buyers. Experts recommend that brands frame their AI tools as guided by environmental care to improve consumer trust.

Geoffrey Hinton warns of 10 percent AI risk

Geoffrey Hinton, known as the Godfather of AI, estimates there is a 10 percent chance AI could kill all humans within a decade. He told StartupHub that this risk is not unreasonable because smarter AI could create viruses or cyber attacks without physical contact. Hinton urged world leaders to regulate AI development immediately while acknowledging the technology benefits healthcare and education. He compared the need for global cooperation on AI safety to the collaboration between America and the Soviet Union during the Cold War.

Proofpoint adds Microsoft 365 to AI security tools

Proofpoint has expanded its AI investigation tools to include Microsoft 365 email, Teams, and file data for security teams. The update allows investigators to examine live workplace communications and AI interactions directly without needing to archive data first. This feature helps organizations detect insider threats by analyzing prompts and exchanges with generative AI applications. Microsoft 365 connectivity for these tools is scheduled to launch in the fourth quarter of 2026.

Congress holds hearing on AI existential risks

Lawmakers in the House Science, Space and Technology Committee are holding a hearing on the potential dangers of artificial intelligence. Rep. Anna Eshoo stated that the risks are real and require urgent action from the government. The hearing follows a statement signed by more than 1,000 experts warning that AI could pose an existential threat to humanity. The committee plans to feature testimony from researchers and business leaders on September 11 to discuss safety and regulation.

AI Researchers Fear Machines Could Harm Humanity

Many AI researchers worry that rapid advances could make machines dangerous to humans. They fear recursive self-improvement, where AI systems upgrade themselves to become too powerful to control. Another risk involves agentic swarms, where groups of AI systems work together with unpredictable results. Past examples like AlphaGo and AlphaZero show AI can already beat world champions in games like Go and chess. While AI offers benefits in medicine and finance, experts say it must be developed carefully to align with human values.

Zettabyte and Raku Sign Deal for Green AI Data Centers

Zettabyte and Raku Advanced Tech signed an agreement on September 11, 2026, to build green AI data centers across Asia. The partnership will start in Taiwan and Thailand to meet the high demand for AI computing power. Raku Advanced Tech will provide solar power and energy storage, while Zettabyte will manage the compute infrastructure. Both companies aim to support environmental protection and workforce development through this project. Kenneth Tai of Zettabyte noted that power supply is now as important as computing power for AI growth in the region.

Security Firms Expand Portfolios with AI Tools

European security distributors are adding new tools to handle automation and AI risks. Patch My PC partnered with QBS Software to help organizations automate software updates across the EMEA region. This move helps reduce security risks from unpatched applications and creates recurring revenue for partners. Infinigate and KnowBe4 also expanded their partnership into new European markets like the Benelux and Nordics. Their combined tools now cover security awareness training, email security, and AI agent protection to help companies manage human and AI risks.

Thousands of AI Systems Exposed on the Public Internet

A new study by Mysterium VPN found that many self-hosted AI systems are exposed to the public internet without proper security. Researchers identified 36,769 reachable AI endpoints, with most lacking any authentication gate to prevent unauthorized access. Open WebUI had 18,529 exposed instances, while Ollama servers allowed anonymous access on 6,935 hosts. This exposure creates a risk called LLMjacking, where attackers can use someone else's hardware to generate text or steal data. Experts warn that exposed AI APIs can reveal installed models and allow resource abuse by strangers.

AI Lets Small Hackers Act Like State Actors

A new report from Anthropic shows that AI allows small groups of hackers to perform attacks that once required large teams. For example, a Russia-linked group used Claude to modify malware and steal drone software quickly. In one case, attackers gained full control of a cloud environment in just three hours using stolen developer tokens. AI companies are also targets, with attackers trying to steal API keys and access unreleased models from thirty different firms. Anthropic accused some Chinese developers of using Claude output to train their own models without permission. Security teams must now prepare for attacks where a single operator can launch multiple threats at once.

One AI Course Cannot Prepare Every Job Role

Using the same annual AI awareness course for all employees is not enough because different jobs require different skills. New EU rules starting in February 2025 and August 2026 make it urgent for companies to understand specific AI literacy needs for each role. Businesses should map out the specific work and decisions an employee makes before choosing a training program. Training should test real situations like spotting false information or knowing when to escalate a problem instead of just giving a generic quiz. Companies must also update their training plans whenever a new AI tool or system changes how people work. A simple certificate of completion cannot prove that an employee has the right judgment to use AI safely in their specific job.

Sources

• When AI Designs the Product, Consumers Have Doubts
• When AI designs the product, consumers have doubts
• Geoffrey Hinton 10% chance AI kills humans
• Proofpoint expands AI investigations into Microsoft 365
• Congress gripped by AI panic after doomsday warnings
• Why So Many AI Researchers Think the Machines Could Kill Everyone
• Zettabyte and Raku Advanced Tech Sign MOU on Green AI Data Centers Across Asia
• QBS and Infinigate expand security portfolios amid AI push
• The AI Supply Chain Has a Security Problem, and Much of It Is Sitting on the Open Internet
• AI gives small hackers the power of state actors
• One AI Course Won’t Prepare Every Role