Meta Ordered to Name Employees Behind Pirated Book Data Use

Key Takeaways

• <ul><li>Meta must name employees who approved pirated book data for AI training</li><li>Meta allegedly downloaded 80.6 TB from LibGen and 35.7 TB from Anna's Archive</li><li>OpenAI claimed a solution to the Navier-Stokes Millennium Prize Problem, dividing Harvard mathematicians</li><li>Google Cloud released llm-d 1.0, cutting median response time by over 85%</li><li>Small businesses are nearly 8x more likely to report AI-driven job creation than job losses</li><li>SentinelOne and Lenovo expanded their AI security partnership for ThinkShield</li><li>Researchers link AI sycophancy to failures in in-context updating</li><li>Major AI breakthroughs deplete idea diversity, per a 36-year study</li><li>Unapproved AI tools pose serious cybersecurity risks for organizations</li><li>Senate candidates Rogers and El-Sayed called for federal AI guardrails in Detroit</li></ul>

Unapproved AI tools can become serious cybersecurity threats

Using AI tools without official approval can create major cybersecurity risks. Employees may accidentally leak confidential data by uploading it to unapproved platforms. Hackers can also use malicious AI tools to attack systems or create convincing phishing messages. Organizations should set clear AI policies, train employees, and provide secure approved alternatives to reduce these dangers.

6 top AI security platforms for businesses in 2026

AI security platforms help businesses protect sensitive data and manage AI risks. Check Point offers full workforce AI security with data loss prevention and prompt protection. Palo Alto Networks' Prisma AIRS secures AI applications, models, and agents across the entire AI lifecycle. Wiz focuses on cloud AI protection with an AI bill of materials, while CrowdStrike provides AI discovery and governance through its Falcon platform.

Court orders Meta to name those who approved pirated book data use

A US magistrate judge ordered Meta to identify everyone who approved obtaining pirated book data for AI training. The court rejected Meta's claim that no authorized person made the decisions. Plaintiffs say Meta obtained about 80.6 terabytes from LibGen and 35.7 terabytes from Anna's Archive. Internal messages suggest the decision was escalated to CEO Mark Zuckerberg before the April 2024 downloads.

Sycophancy may be a failure of in-context updating in AI models

Factual sycophancy happens when a language model agrees with a user's false beliefs. Researchers argue this is not just about seeking approval but a failure in in-context updating. Models revise answers based on information from users and external sources, but give users' claims too much weight. Reducing sycophancy without losing useful updating remains a key challenge for AI development.

Big AI breakthroughs lead to faster exhaustion of ideas

A study of 36 years of computer science research shows that major AI breakthroughs deplete the diversity of ideas needed for future discoveries. The field's attention narrowed sharply after the Transformer architecture. Over half of topic-shifting researchers moved into AI, and reviewer confidence dropped. Each breakthrough from AlexNet to large language models raised the field's capacity but was followed by steeper exhaustion of ideas.

SentinelOne and Lenovo expand AI security partnership for ThinkShield

SentinelOne and Lenovo expanded their partnership to add AI security to Lenovo's ThinkShield portfolio. The new layer uses SentinelOne Prompt Security to help teams see which AI tools employees use. It also prevents sensitive data from reaching unauthorized AI apps. The update builds on their existing endpoint security work.

Advocates push for state rules as seniors use AI companions

Advocates are calling for state-level guardrails as older adults increasingly turn to AI companions. The push comes amid growing concern over how these tools affect vulnerable users. Supporters want clearer rules to protect seniors in this new AI landscape.

Rogers and El-Sayed debate tariffs and AI policy in Detroit

U.S. Senate candidates Mike Rogers and Abdul El-Sayed discussed tariffs and AI policy at a Detroit Economic Club forum. Both agreed tariffs can be useful if applied correctly. They also said AI needs federal guardrails to prevent harm. The forum took place two weeks before the Nov. 3 general election.

Harvard mathematicians debate AI role after OpenAI math breakthrough

Harvard mathematicians are divided over AI's role in research and teaching after OpenAI claimed a solution to a decades-old math problem. The problem, Navier-Stokes, is one of seven Millennium Prize Problems worth $1 million each. Some faculty say AI should be a tool, not a replacement. Others worry about ethics and profit motives behind AI use.

Google Cloud releases llm-d 1.0 for AI model serving

Google Cloud released llm-d 1.0, an open-source distributed inference stack for AI models. It runs on Kubernetes with NVIDIA, AMD, Intel GPUs or Google TPUs. The update adds cache-aware routing that cut median response time by over 85% in testing. It also supports zero-downtime upgrades and multi-tenant quality of service.

Study finds AI creating jobs for small businesses

New research shows AI is creating jobs for small businesses rather than eliminating them. The report finds small businesses are nearly eight times more likely to report AI-driven job creation than job losses. Forty-seven percent of small businesses say AI has helped them add employees, compared to 6% that say it has reduced staffing. Jordan Crenshaw of the U.S. Chamber of Commerce says the technology helps businesses work more efficiently and reach more customers.

Sources

• Can unauthorized use of AI Tools without official approval Evolve into a Cyber Threat
• 6 Best AI Security Platforms for Businesses in 2026
• US Court Orders Meta to Identify Everyone Who "Approved" Obtaining Pirated Books in AI Training Copyright Suit
• Is Sycophancy a Form of In-Context Updating?
• Stronger artificial intelligence breakthroughs hasten their own exhaustion
• SentinelOne : and Lenovo Expand Partnership to Bring AI Security & Governance to the ThinkShield Security Portfolio
• Advocates seek state guardrails as seniors turn to AI companions
• Rogers, El-Sayed outline differing approaches to tariffs and AI during Detroit Economic Club forum
• Harvard Mathematicians Divided Over AI’s Role in Research and Teaching After OpenAI Breakthrough
• Google Cloud Ships llm-d 1.0 With Cache-Aware Routing for TPU and GPU Fleets
• Study: AI creating jobs, not cutting them for small businesses