The Financial Stability Board has released 12 sound practices aimed at guiding banks and financial institutions in responsible AI use. The practices address governance, risk management, data oversight, human control, and cybersecurity, with an emphasis on tailoring each institution's approach. The FSB also encourages cross-border coordination and sharing of AI-related insights among financial regulators.
In a separate report, the FSB urged banks to adopt company-wide regulation for agentic AI systems. The warning highlights that AI can amplify risks such as poor transparency, unpredictable actions, and faster cyber attacks. The board also noted that failing to adopt AI can reduce competitiveness and weaken threat detection capabilities across the financial sector.
An Arizona appeals court ruling could shape how courts nationwide handle AI-generated evidence. The court decided that an AI-generated video of a shooting victim should not be used in sentencing, finding it made the hearing unfair by presenting imagined statements as if they came from the victim. Experts say the decision may set a precedent since few states have addressed this issue directly.
AI is making traditional scams cheaper and harder to detect. Voice clones now cost as little as one to 20 cents per minute, and the FBI logged $893 million in AI-related fraud complaints in 2025. Experts warn that familiar warning signs like misspellings or awkward phrasing are no longer reliable indicators, so defenses must move earlier in the fraud detection process.
A new study reveals AI writing is appearing in more dissertations across academic fields. AI writing was absent before 2023 but showed up in 29% of STEM dissertations filed in 2026. In philosophy, AI writing rose from 1.3% in 2024 to 7.8% in 2025. Students who rely on AI in their dissertations are less likely to pursue academic research careers, raising concerns about the development of scientific expertise.
Latent risk is emerging as a significant concern as AI systems handle more tasks over extended periods. Workers who monitor AI may lose the skills needed to perform tasks independently. Companies like Deloitte are adjusting training methods, and experts say organizations must keep testing human skills and update accountability rules when AI fails.
Meta and Apple collaborated on a new AI model called Muse Spark Open. The model has 816 billion parameters and runs efficiently on M5 Ultra Mac Studios. Meta CEO Mark Zuckerberg and Apple CEO John Ternus both supported the project, which is designed for long tasks, complex coding, and multi-step reasoning.
AI tools like Anthropic's Claude Opus 5.5 are fueling a rise in vibe-coded game mods, according to Take-Two's former AI head Luke Dicken. Game developers warn the trend will lead to stronger DRM and streaming-based games. Publishers are concerned that AI can decompile source code faster than they can protect it.
At a Workday event in Las Vegas, Intermountain Health and BMO demonstrated real-world AI applications. Intermountain Health tested agents like Developer Agent and Payroll Agent to support caregivers, while BMO gave Workday's self-service agent to all 55,000 employees in June. Both companies emphasized managing risks while improving task efficiency.
Arkansas candidates for the 3rd Congressional District debated AI regulation and data center rules. Republican Steve Womack, Democrat Robb Ryerse, and Libertarian Bobby Wilson spoke at an October 14, 2026 debate, disagreeing on AI use in warfare. Womack supports drone maker Swarm Aero, which holds a contract with the Air Force Research Laboratory.
Sociologist Sherry Turkle appeared on The Ezra Klein Show to discuss emotional connections with AI. She argues that caring for an AI resembles entering a shell game and is not as genuine as caring for a pet. The conversation explored the complexities of human relationships with increasingly sophisticated technology.
Key Takeaways
- The FSB finalized 12 practices for responsible AI use in finance, covering governance, risk management, data oversight, human control, and cybersecurity.
- The FSB warned that failing to adopt AI can reduce competitiveness and weaken threat detection in banks.
- An Arizona appeals court ruled an AI-generated video of a shooting victim cannot be used in sentencing, potentially setting a nationwide precedent.
- The FBI logged $893 million in AI-related fraud complaints in 2025, with voice clones costing as little as one to 20 cents per minute.
- AI writing appeared in 29% of STEM dissertations filed in 2026, up from zero before 2023.
- AI writing in philosophy dissertations rose from 1.3% in 2024 to 7.8% in 2025, with users less likely to pursue academic careers.
- Latent risk is growing as workers who monitor AI systems may lose the ability to perform tasks without assistance.
- Meta and Apple developed Muse Spark Open, an 816-billion-parameter AI model optimized for M5 Ultra Mac Studios.
- Anthropic's Claude Opus 5.5 is driving a rise in vibe-coded game mods, raising concerns about source code decompilation and DRM.
- BMO deployed Workday's self-service agent to all 55,000 employees in June, while Intermountain Health tested Developer and Payroll Agents for caregivers.
FSB releases 12 practices for responsible AI use in finance
The Financial Stability Board finalized 12 sound practices to guide banks and financial institutions in using AI responsibly. The practices cover governance, risk management, data oversight, human control, and cybersecurity. They are designed to be tailored to each institution's needs. The report also encourages sharing and coordination across borders.
FSB urges banks to adopt company-wide AI regulation
The Financial Stability Board called on banks to increase oversight of agentic AI in a new report. It warns that AI can amplify risks like poor transparency, unpredictable actions, and faster cyber attacks. The report also says failing to adopt AI can reduce competitiveness and weaken threat detection. Earlier guidance treating AI as agents rather than tools was also referenced.
Arizona AI ruling may shape how courts handle AI evidence
An Arizona appeals court ruled that an AI-generated video of a shooting victim should not be used in sentencing. The court found the video made the hearing unfair by presenting imagined statements as if they came from the victim himself. Experts say the ruling could influence courts nationwide since few states have addressed this issue. Defense attorneys see potential for AI to help victims who can no longer speak, but worry about fabricated statements.
AI makes old scams cheaper and harder to spot
AI has not created new scams but has made impersonation fraud much cheaper and more convincing. Voice clones now cost as little as one to 20 cents per minute, and the FBI logged $893 million in AI-related fraud complaints in 2025. The FBI logged $893 million in AI-related fraud complaints in 2025. Experts say traditional warning signs like misspellings or awkward phrasing are no longer reliable, so defenses must move earlier in the process.
AI writing appears in more dissertations across fields
A new study shows AI writing was absent in dissertations before 2023 but appeared in 29% of dissertations filed in 2026 in STEM fields. In philosophy, AI writing rose from 1.3% in 2024 to 7.8% in 2025 and is still growing. Students using AI in dissertations are less likely to pursue academic research careers. The study raises concerns about how AI use may reduce the development of scientific expertise.
As AI handles more work, humans may lose skill at stepping in
AI systems that work well for months can cause a hidden risk. The person watching the AI may forget how to do the task without help. This problem is called latent risk. Companies like Deloitte are changing how they train workers. Experts say organizations must keep testing human skills and update rules about who is responsible when AI fails.
Meta and Apple team up on Muse AI model for Macs
Meta and Apple worked together on a new AI model called Muse Spark Open. The model has 816 billion parameters and runs well on M5 Ultra Mac Studios. Meta CEO Mark Zuckerberg and Apple CEO John Ternus both supported the project. The model is built to handle long tasks, complex coding, and multi-step reasoning.
Vibe coding is changing the video game industry
AI tools like Anthropic's Claude Opus 5.5 are causing a rise in vibe-coded game mods. Take-Two's former AI head Luke Dicken says this creates serious problems for game makers. Godot's Emilio Coppola also warns the trend will lead to stronger DRM and streaming-based games. Publishers fear AI can decompile source code faster than they can protect it.
Workday customers show how to use AI in real businesses
At a Workday event in Las Vegas, Intermountain Health and BMO shared how they use AI. Intermountain Health tested agents like Developer Agent and Payroll Agent to help caregivers. BMO gave Workday's self-service agent to all 55,000 employees in June. Both companies focused on using AI to make tasks easier while managing risks.
Arkansas candidates debate AI rules and data centers
Candidates for Arkansas' 3rd Congressional District agreed AI needs regulation. Republican Steve Womack, Democrat Robb Ryerse, and Libertarian Bobby Wilson spoke at a debate on October 14, 2026. They disagreed on AI use in warfare and data center rules. Womack supports drone maker Swarm Aero, which has a contract with the Air Force Research Laboratory.
Sociologist Sherry Turkle discusses AI care on Ezra Klein Show
Sociologist Sherry Turkle appeared on The Ezra Klein Show to discuss the relationship between humans and artificial intelligence. She argues that caring for an AI is similar to entering a shell game. Turkle suggests this relationship is not as genuine as caring for a pet. The conversation explores the complexities of emotional connections with technology.
Sources
- FSB finalises sound practices for the responsible adoption of artificial intelligence (AI)
- FSB Says Banks Need to Have Company-Wide AI Regulation
- Arizona AI ruling could influence courts nationwide
- AI didn't invent new scams. It made the old ones cheap.
- AI Use in STEM and Philosophy Dissertations
- As AI takes on more tasks, what happens when people have to jump in to clean up after it?
- Meta and Apple team up to make Muse AI models run better on Macs
- Interview: The Video Game Industry’s Vibe-Coded Future
- Workday customers showcase how to unlock AI value
- Northwest Arkansas candidates tackle AI, affordability in US House debate
- Video: Opinion | Can A.I. Care for You?
Comments
Please log in to post a comment.