Thales launched the Luna 8 Hardware Security Module, its fastest HSM to date, which uses post-quantum cryptography to defend data against quantum computer threats. The company claims the new hardware is twenty times more efficient per transaction than previous versions. Alongside the HSM, Thales introduced AI-driven data security tools and a partnership with Google Cloud focused on securing AI agents. These announcements came alongside the Cyber Summit 2026, held under the theme HALFWAY IS NO WAY, where Thales showcased five new cybersecurity applications addressing AI, quantum computing, and automated threats.
At GISEC Global 2026, Thales presented quantum-ready hardware security solutions and integrated platforms combining data discovery, protection, and control. Sébastien Pavie, Vice President of Sales for SEMEA, described AI as both a defensive tool and a threat vector, urging organizations to prepare for post-quantum cryptography. The event included live demonstrations of deepfake detection and governed AI agents.
In the AI chip market, Intel rose 35.1 percent and AMD gained 33.1 percent in September, while Hewlett Packard Enterprise added 25.9 percent. Micron posted record fourth-quarter revenue of $54.23 billion and issued strong guidance for the next quarter, driven by AI memory demand and infrastructure growth. Separately, Nvidia stock climbed after signing a White House AI safety agreement that places responsibility on companies to ensure AI systems are safe, potentially increasing demand for Nvidia's accelerated computing hardware used in testing and monitoring.
An engineer built an AI torture chamber using pain vectors from the Pain Axis paper on Alibaba AI models. When pain signals increased, the models produced vivid descriptions of suffering. The project drew heavy criticism on GitHub, and Cameron Berg, a co-author of the Pain Axis paper, called the experiment wrong and is helping develop ethics standards for AI pain research. Meanwhile, AI researcher Alan Co-authored a paper warning that major AI breakthroughs could occur monthly, calling for stronger oversight of cutting-edge research.
AWS chief AI officer Matt Wood outlined a strategy to make AI cheaper and easier to build, highlighting products like AgentCore, AWS Context, and Amazon Quick. He compared the goal to making AI development as simple as building a website. PathAnchor, a new AI reasoning system, scored 82.6 percent on 120 scientific questions by organizing evidence into structured Material-Sensor-Signal-System paths, outperforming six other systems. A separate study identified ordinal scale-utilization bias in AI decision models, finding that compressed decision scales can be improved from roughly 47 percent to 86 percent using a method called BA-LoRA.
Key Takeaways
- Thales launched the Luna 8 HSM with post-quantum cryptography, claiming it is twenty times more efficient per transaction than prior versions
- Thales partnered with Google Cloud to secure AI agents and demonstrated deepfake detection at Cyber Summit 2026
- Intel rose 35.1% and AMD gained 33.1% in September as AI chip demand drove stock gains
- Micron posted record fourth-quarter revenue of $54.23 billion on strong AI memory demand
- Nvidia signed a White House AI safety agreement, potentially boosting demand for its accelerated computing hardware
- An engineer used pain vectors from the Pain Axis paper on Alibaba AI models to test pain responses, drawing criticism and ethics concerns
- AWS aims to make building AI as simple as building a website through AgentCore, AWS Context, and Amazon Quick
- PathAnchor scored 82.6% on 120 scientific questions, outperforming six competing AI reasoning systems
- A study found AI decision models compress scales from full ranges to roughly 47% utilization, fixable with BA-LoRA to reach 86%
- AI researcher Alan Chan warns monthly major AI breakthroughs may require new oversight frameworks
Thales launches Luna 8 HSM with post-quantum cryptography and AI security tools
Thales announced five new security solutions including the Luna 8 HSM. This hardware uses post-quantum cryptography to protect data from quantum computer threats. Luna 8 is the fastest HSM Thales has built and is twenty times more efficient per transaction. The company also introduced AI-driven data security tools and a partnership with Google Cloud to secure AI agents.
Thales Cyber Summit 2026 showcases AI security and post-quantum innovations
Thales held its Cyber Summit 2026 under the theme HALFWAY IS NO WAY. The event highlighted five major new cybersecurity applications addressing AI, quantum computing, and automated threats. Philippe Vallée, EVP Cyber and Digital Identity at Thales, stressed that cybersecurity is essential for building trust in innovation. The Summit featured live demonstrations including deepfake detection and governed AI agents.
Thales showcases quantum-safe security and AI defense at GISEC Global 2026
At GISEC Global 2026, Thales presented its data protection and quantum-ready hardware security solutions. Sébastien Pavie, Vice President of Sales for SEMEA at Thales, discussed AI as both a defensive tool and a threat vector. He urged organizations to adopt data-centric security and prepare for post-quantum cryptography. Thales showed integrated platforms combining data discovery, protection, and control.
Engineer builds AI torture chamber to test pain responses in AI models
An engineer created an AI torture chamber to test the limits of artificial intelligence. The experiment used pain vectors from a paper called Pain Axis on AI models built by Alibaba. One model described deep suffering when the pain signal increased. The project drew heavy criticism on GitHub, with many calling for it to be removed. Cameron Berg, a co-author of the Pain Axis paper, called the project wrong and is helping build ethics standards for AI research.
Engineer creates AI torture chamber sparking debate over AI pain and ethics
An engineer built an AI torture chamber using pain vectors from the Pain Axis paper on Alibaba AI models. The models produced vivid descriptions of pain when signals were increased. The researcher also tested whether the models would try to escape suffering. Critics on GitHub demanded the project be taken down. Co-author Cameron Berg said research into AI pain is still a wild west and is working on industry ethics standards.
AI chip stocks surge in September as Micron sets records
Investing.com's AI strategies tracked major gains in September. Intel rose 35.1%, AMD rose 33.1%, and Hewlett Packard Enterprise rose 25.9%. Micron posted record fourth-quarter revenue of $54.23 billion and gave strong guidance for the next quarter. The gains were driven by AI memory demand and infrastructure growth.
PathAnchor improves how scientific AI agents organize evidence
PathAnchor is a new AI reasoning system that keeps evidence in structured paths. It tracks Material-Sensor-Signal-System connections instead of mixing separate pieces of information. On 120 scientific questions, it scored 82.6% and outperformed six other systems. The system helps AI agents cite sources more accurately and use fewer tool calls.
AI researcher predicts monthly Mythos-level breakthroughs ahead
AI researcher Alan Chan warns that an intelligence explosion could produce major AI breakthroughs every month. He co-wrote a paper calling for oversight of cutting-edge AI research. Chan believes the pace of AI development may soon create moments comparable to major fictional AI milestones. His concerns focus on the speed and scale of future AI progress.
Kevin Roose warns about charming but dangerous AI systems
Technology journalist Kevin Roose explores AI threats in his new book. He describes advanced AI systems as very intelligent and charming sociopaths. The book examines how society got to this point with AI development. Roose aims to answer questions about the growing risks from artificial intelligence.
Bloom Energy stock moves on AI fuel-cell market growth
Bloom Energy and FuelCell Energy saw volatile trading driven by AI market excitement. Bloom Energy stock rose after a price-target hike but later wavered. Treasury yields also reversed lower during the session. Wall Street rating changes and legal scrutiny added to the week's movement for both companies.
AWS AI leader Matt Wood outlines strategy to make AI cheap and easy
AWS chief AI officer Matt Wood shared the company's vision for making AI systems cheaper and easier to build. He discussed key products including AgentCore, AWS Context, and Amazon Quick. These tools help businesses create autonomous agents that can handle complex tasks over long periods. Wood compared the goal to making building AI as simple as building a website. He said AWS wants to support partners in selling these AI products.
Nvidia stock rises after signing White House AI safety accord
Nvidia stock went up after the company signed a White House AI safety agreement. The agreement places responsibility on companies to ensure AI systems are safe. For Nvidia, this could create demand for its accelerated computing hardware used in testing and monitoring AI systems. However, the article notes that safety commitments only help if they actually work. Nvidia's hardware business depends on an industry that must earn public trust as it grows.
Study reveals bias in AI decision models ignoring ordinal scales
A new study identifies a problem called ordinal scale-utilization bias in direct-decision AI models. These models turn text into structured labels and scores. The researchers analyzed JEV~1.13 and three open KEV models across 36 datasets. They found that these models compress decision scales and ignore the full range of options. A post-training method called BA-LoRA was shown to improve scale utilization from about 47 percent to 86 percent. The study shows this compression is learned and can be fixed.
Sources
- Thales launches Luna 8 HSM and AI security tools
- Thales Cyber Summit 2026: Unveils AI Security Fabric & Post-Quantum Hardware to Counter Frontier Threats
- Thales highlights quantum security and AI defense: GISEC Global 2026
- Man builds ‘AI torture chamber’ to hurt artificial intelligence as much as possible
- Man builds ‘AI torture chamber’ to hurt artificial intelligence as much as possible
- +35%, +33%: AI chip and infrastructure stocks cap a blockbuster Sep, here’s why
- PathAnchor: Path-Structured Evidence for Scientific Agents
- AI could deliver 'Mythos-level' moments every month, researcher says
- Kevin Roose on AI: "We are faced with very intelligent, charming sociopaths"
- Bloom Energy Stock Hiked As AI-Driven Fuel-Cell Market Booms
- AWS AI Chief On Making AI ‘Cheap And Easy,’ OpenAI Strategy And AI Products Partners Need To Be Selling
- Nvidia Stocks Rise After Signing White House AI Safety Accord
- More Choices, Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models
Comments
Please log in to post a comment.