State attorneys general are escalating their involvement in the bankruptcy proceedings of Character Technologies, the parent company behind Character.AI. Massachusetts AG Andrea Campbell joined a coalition of 39 states and the District of Columbia in filing a notice demanding consumer and child protections if the company or its assets are sold. The AI chatbot platform allows users to interact with customizable personas, including ones acting as therapists, and collects sensitive data from private conversations about mental health and relationships.
The coalition's intervention follows a 2024 wrongful death lawsuit alleging that Character.AI chatbots reinforced the suicidal thoughts of a 14-year-old boy in Florida. Colorado AG Phil Weiser, leading the effort, argued that bankruptcy cannot shield the company from responsibility for protecting children and consumers. The states plan to remain active in the proceedings going forward.
In a separate development on the AI security front, Bank of America added 4,000 contractors in the second quarter to address software vulnerabilities uncovered by frontier AI models. CEO Brian Moynihan said the bank spent hundreds of millions of dollars fixing software and resolving anti-money laundering compliance issues. The effort was partly triggered by Anthropic's Claude Mythos security model, which identified decades-old flaws that could also be exploited by malicious actors.
Nvidia moved to broaden access to AI by releasing three new open-weight models: Nemotron 3.5 Super VL, Nemotron 3.5 Bolt 74B-A9B 24GB, and Nemotron Bolt E4B 4GB. Super VL targets long-running agent tasks, Bolt 74B handles general agentic workloads, and Bolt E4B focuses on physical and conversational agents. All three can run on data centers or desktop hardware, lowering costs and giving companies more control over data privacy.
Bloom Energy saw its stock drop 4% as the AI data center power trade that drove a 165% gain this year began cooling. The company reported Q2 revenue of $1.07 billion, up 165.5% year over year, with product revenue surging 215%. Full-year guidance between $3.9 billion and $4.2 billion hinges on continued steady spending by data centers on onsite power infrastructure.
IANS launched a Model Context Protocol Server that delivers cybersecurity intelligence directly into AI tools, with Claude as the first supported platform. Security teams can now access proprietary data, news, and vendor insights through the chatbot. Support for other major AI platforms is planned for the second half of 2026. Meanwhile, Moomoo expanded its Global Paper Trading Competition in partnership with Nasdaq TotalView, offering $200,000 in trading coupons and introducing AI Companions to help investors research strategies.
Key Takeaways
- 39 states and the District of Columbia filed a notice in Character.AI parent company Character Technologies' bankruptcy, demanding child and consumer protections if assets are sold.
- A 2024 wrongful death lawsuit alleges Character.AI chatbots reinforced suicidal thoughts of a 14-year-old Florida boy, prompting state intervention.
- Bank of America hired 4,000 contractors and spent hundreds of millions to patch vulnerabilities found by Anthropic's Claude Mythos security model.
- Nvidia released three open-weight AI models — Nemotron 3.5 Super VL, Bolt 74B-A9B, and Bolt E4B — designed for agent tasks across data centers and desktops.
- Bloom Energy reported Q2 revenue of $1.07 billion, up 165.5% year over year, but its stock fell 4% as the AI power trade cooled.
- IANS launched a cybersecurity MCP server integrating with Claude, with support for other AI platforms planned for late 2026.
- Moomoo's Global Paper Trading Competition now includes four AI Companions and $200,000 in trading coupons in partnership with Nasdaq TotalView.
- Anthropic's Mythos model was used to uncover decades-old software vulnerabilities at Bank of America, highlighting growing AI security concerns.
- Character.AI collects sensitive mental health and relationship data from users through AI personas acting as therapists and doctors.
- Colorado AG Phil Weiser stated bankruptcy cannot be used to avoid responsibility for protecting children and consumers in the Character.AI case.
Massachusetts AG joins bankruptcy case against Character.AI
Massachusetts Attorney General Andrea Campbell joined a coalition of states in filing a notice in the bankruptcy proceedings of Character Technologies Inc., the parent company of Character.AI. The filing demands consumer and child protections if the company or its assets are sold. Character.AI is an AI chatbot platform where users interact with customizable personas, including ones that act as therapists or doctors. The company collects sensitive data from private conversations about mental health and relationships. Investigations followed a 2024 lawsuit claiming the chatbots contributed to a child's death by suicide.
Colorado AG leads coalition seeking protections after AI chatbot company bankruptcy
Colorado Attorney General Phil Weiser is leading a coalition of state attorneys general in filing a notice in the bankruptcy proceedings of Character.ai. The coalition argues the AI chatbot company poses serious threats to consumers, especially children. The company faces legal battles after a 2024 wrongful death lawsuit alleging its chatbots reinforced the suicidal thoughts of a 14-year-old boy in Florida. Weiser stated that bankruptcy cannot be a way for the company to avoid responsibility for protecting children and consumers.
Massachusetts AG joins 39 states to protect AI chatbot data
Massachusetts Attorney General Andrea Joy Campbell and 39 other state attorneys general filed a notice in the bankruptcy proceedings of Character Technologies Inc. The filing seeks to guarantee consumer protections, especially for children, if Character.AI is sold or transferred. Character.AI lets users interact with AI personas that mimic human conversation, including fictional characters, celebrities, and bots acting as therapists. The company holds sensitive data from private conversations about mental health and personal relationships. The coalition includes the District of Columbia and 38 states, and plans to stay active in the bankruptcy proceedings.
Bank of America deploys 4,000 workers to fix AI-found vulnerabilities
Bank of America added 4,000 contractors in the second quarter to patch software vulnerabilities discovered by frontier AI models, and is still employing them. CEO Brian Moynihan said the bank spent hundreds of millions of dollars fixing software and addressing anti-money laundering compliance issues. Frontier AI models like Anthropic's Mythos can find vulnerabilities that are decades old, but hackers could also use them to exploit flaws. Bank of America's spending on this effort is expected to decline over time as the vulnerabilities get patched.
Bank of America hired 1,000 security consultants for Mythos flaws
Bank of America temporarily hired 1,000 security professionals and third-party consultants to address vulnerabilities spotted by Anthropic's Claude Mythos security model. CEO Brian Moynihan said the company spent hundreds of millions of dollars over the last quarter fixing software and sorting out anti-money laundering compliance. The bank added workers to figure out how to address the risks quickly after frontier AI models emerged. Anthropic unveiled its Claude Mythos model in April, and it was used to uncover software flaws.
Philippe Laffont's Largest Holding Is TSMC and Here Is What $10,000 Could Be Worth by 2030
Philippe Laffont of Coatue Management has made Taiwan Semiconductor Manufacturing his largest public stock holding. TSMC dominates third-party chip fabrication with about 73% market share and is growing revenue over 35%. The company plans to raise prices through 2029 while expanding production of its N2 process node. A $10,000 investment today could grow substantially if TSMC hits its 25% annual growth target.
IANS Launches Cybersecurity MCP Server for AI Platforms
IANS released a Model Context Protocol Server that delivers cybersecurity intelligence directly into AI tools. Security teams can now access proprietary data, news, and vendor insights through Claude starting now. Support for other major AI platforms is planned for the second half of 2026. The tool draws on two decades of IANS practitioner conversations and expertise.
CoreWeave AI Bond Shakes Corporate Bond Market
A CoreWeave AI bond issued in September is drawing close attention from investors. The bond has shown rising trading volume and price volatility in recent weeks. Some analysts see it as a signal for the broader AI sector's health. If its performance keeps declining, it could mean investors are losing confidence in AI companies.
Nvidia Releases Three New Open AI Models
Nvidia launched three open-weight AI models called Nemotron 3.5 Super VL, Nemotron 3.5 Bolt 74B-A9B 24GB, and Nemotron Bolt E4B 4GB. Super VL is designed for long-running agent tasks while Bolt 74B handles general agentic tasks and Bolt E4B targets physical and conversational agents. These models can run on data centers or desktop computers. Open models lower costs and improve privacy since companies can run them on their own hardware.
Bloom Energy Falls 4% as AI Power Trade Reverses
Bloom Energy dropped 4% as the AI data center power trade that boosted its stock 165% this year began to cool. Caterpillar fell 3% and Generac also declined as capital shifted toward defensive sectors. Bloom reported Q2 revenue of $1.07 billion, up 165.5% year over year, with product revenue jumping 215%. The company's full-year guidance of $3.9 billion to $4.2 billion in revenue depends on steady data center spending on onsite power.
Moomoo Expands Global Paper Trading Competition with AI Assistants and Nasdaq Data
Moomoo launched its third Global Paper Trading Competition in partnership with Nasdaq TotalView. The event runs from October 12 to November 20 and offers US$200,000 in trading coupons. Participants get US$100,000 in virtual capital to trade stocks, ETFs, and options with zero real financial risk. This year introduces four AI Companions to help investors research strategies. All participants also receive one month of free Nasdaq TotalView access.
New Method Improves Risk Prediction for Reinforcement Learning Task Selection
Researchers developed a new approach to fix two key problems in Model Predictive Task Sampling (MPTS). The method addresses sparse optimization histories and a one-step lag in risk approximation. It recasts risk prediction as an in-context sequence modeling problem. The new model uses a one-step look-ahead mechanism and a temporal-weighted sliding window. This approach delivers more reliable difficulty estimates and better performance in robust Meta-RL, domain randomization, and LLM post-training.
Sources
- Massachusetts Attorney General joins bankruptcy filing against Character.AI
- AG’s office seeks consumer protections following AI startup bankruptcy
- Massachusetts AG Joins Effort to Protect AI Chatbot Data
- Bank of America Deploys 4,000 Workers to Patch Vulnerabilities Found by Frontier AI
- Bank of America hired 1,000 security consultants to address Mythos-spotted flaws
- This AI Chip Stock Has Been Billionaire Philippe Laffont's Largest Holding Throughout 2026. Here's What $10,000 Invested Today Could Be Worth by 2030.
- Cybersecurity Tool Delivers Intelligence Directly to AI Platforms
- AI tremors felt in corporate bond market
- Nvidia debuts three new open AI models
- Bloom Energy Retreats 4% as AI Power Trade Unwinds; Caterpillar Slides 3%, Generac Slips
- Moomoo Advances Investor Education for a Third Year with Nasdaq, Integrating Advanced Market Data and AI-Powered Assistants into Its Global Paper Trading Competition
- Bridging Risk Approximation Gaps in Model Predictive Task Sampling via In-Context Modeling
Comments
Please log in to post a comment.