Bank of Canada and Lenders Discuss Anthropic AI Risks

The Frontier

Science & tech breakthroughs in AI models, hardware, and computing that define what's possible

Bank of Canada, Major Lenders Met on Anthropic AI Cyber Risk. The Bank of Canada and major financial institutions convened to address cybersecurity risks associated with Anthropic's AI model, indicating growing concerns about AI's implications for financial security.

Bank of Canada pulled Canada's Big Six, who hold roughly 90 percent of national deposits, into a same-day briefing on Anthropic's new AI. Rapid, pre-incident coordination last appeared after SolarWinds in 2020. Expect AI-specific cyber rules for systemically important banks before Q3.

Financial Post

ChatGPT rolls out new $100 Pro subscription to challenge Claude. OpenAI introduced a new $100 Pro subscription for ChatGPT, aligning with Claude's pricing strategy, indicating a competitive shift in AI subscription models aimed at enterprise users.

This pricing adjustment reflects OpenAI's strategy to capture a larger share of the enterprise market, potentially reshaping user expectations around AI service costs.

Bleepingcomputer

Anthropic's Claude for Word is another challenge to Microsoft's software empire. Anthropic launched Claude for Word, expanding its AI capabilities to compete directly with Microsoft's software suite, marking a significant shift in the AI productivity tools landscape.

This move could intensify competition in the AI software market, particularly in legal and professional sectors where AI-enhanced document processing is increasingly valuable.

Businessinsider

Model Developments

Latest AI model releases, capabilities, and technical breakthroughs shaping the competitive landscape

AI models are terrible at betting on soccer—especially xAI Grok. A study revealed that AI models from leading companies, including Google and OpenAI, lost money when betting on soccer matches, highlighting limitations in AI's predictive capabilities over extended periods.

The findings suggest that even advanced AI systems struggle with real-world applications, raising questions about their reliability in dynamic environments like sports betting.

Ars Technica

Capital & Control

Tracking the flow of capital and influence that shapes the tech landscape

Motorola acquires AI-based 911 call screening startup Hyper. Motorola Solutions has acquired Hyper, an AI-based startup that screens non-emergency 911 calls, indicating a strategic move to enhance public safety technology through AI integration.

Motorola pre-empts the Series A cycle, snapping up Hyper less than a year out of stealth. The likely sub-$100M exit hands seed backers 10-20× and folds multilingual call-triage AI into Motorola’s 80% US public-safety comms moat, pressuring RapidSOS and AWS voice pilots. Net effect: Hexagon, Carbyne or Verizon to chase comparable startups by Q4 to avoid paying Motorola’s toll.

Betakit

Infrastructure & Power

The physical backbone of AI: compute, data centers, supply chains, and the geopolitics of energy

PJM Targets 15 Gigawatts of New Power for Data Center Boom. PJM Interconnection LLC's proposal for 15 gigawatts of new power supplies addresses potential electricity shortages driven by the surge in artificial intelligence data centers, highlighting the urgent need for infrastructure to support AI growth.

This initiative reflects the increasing demand for energy resources as AI technologies expand, signaling potential shifts in energy policy and investment in power infrastructure.

Financial Post

Digital Defense

Threats and defenses shaping the global security landscape

Nearly 4,000 US industrial devices exposed to Iranian cyberattacks. Iranian state-backed hackers have targeted nearly 4,000 U.S. industrial devices, specifically programmable logic controllers (PLCs) from Rockwell Automation, indicating a significant threat to critical infrastructure.

3,900 internet-facing Rockwell PLCs hand Tehran real-time control knobs across US manufacturing, dwarfing the 1,800 exposed Siemens units Russia scanned in 2022. With OT patch cycles averaging 15 months, Net effect: latent footholds to trigger Q4 supply shocks.

Bleepingcomputer

The Performance Edge

Peak performance science for sustained execution in high-stakes environments

Early weight gain is linked to lifelong health consequences. A study from Lund University reveals that early weight gain significantly increases the risk of premature death from major diseases, emphasizing the long-term health impacts of obesity.

600k-person Swedish cohort shows obesity by 25 sharply raises lifetime death risk from heart disease and diabetes. Risk rose with years carrying extra weight, while female cancer deaths stayed flat. Net effect: BMI tracking to shift to adolescents within 3 years as payers target early prevention.

Science Daily

Crypto & Digital Assets

Bitcoin, Ethereum, DeFi protocols, and digital asset market dynamics

Over 20,000 crypto fraud victims identified in international crackdown. An international law enforcement operation identified over 20,000 victims of cryptocurrency fraud, highlighting the scale of illicit activities in the crypto space and the collaborative efforts to combat them.

20,000 victims with only $12M frozen shows sub-10 % clawback efficiency, boosting demand for on-chain forensics as vendors like Chainalysis ride an $8B valuation. If UK-US sweeps widen, Tier-2 exchanges will exit or sell within 18 months under rising compliance burn.

Bleepingcomputer

The Terminal

Strategic intelligence for builders and operators

name:. Fixed several plugin issues: slash commands resolving to the wrong plugin with duplicate name: frontmatter, /plugin update failing with ENAMETOOLONG, Discover showing already-installed plugins, directory-source plugins loading from a stale version cache, and skills not honoring context: fork and agent frontmatter fields

This update fixes various issues with plugins, making it easier for users to find and use the right tools without confusion or errors. Now, users can smoothly update plugins and access the latest features without running into problems caused by outdated information.

Claude Code (v2.1.101)