The SEC has filed charges against four entities — Cryptoaiml Ltd., Cryptoaiml Capital Foundation, TSAI Pro Ltd., and TSAI Capital Foundation — for allegedly running a $15 million AI trading scam. According to the complaint filed September 29 in Manhattan, the operators used WhatsApp and Facebook groups to lure hundreds of retail investors with promises of AI-powered profits. They falsely claimed SEC oversight and posted fabricated regulatory documents to build trust.
The SEC says the entities misappropriated more than $12.5 million from one group and $2.8 million from another. Between August 2024 and March 2025, the operators convinced investors that large trading gains were real, but the SEC found no actual trading occurred and all displayed profits were fictitious. Investors later faced frozen accounts and were asked to pay additional fees to withdraw funds. The agency recommends using Investor.gov to verify investment providers before committing money.
In other AI-related developments, Robinhood CEO Vlad Tenev announced a suite of new AI-powered trading tools at the HOOD Engines of Creation summit. The company also launched an AI trading agent that operates around the clock, providing real-time market data and analysis. Robinhood shares rose 4.5% in pre-market trading after the announcement, signaling strong investor interest in the company's AI push.
Anthropic released a concerning report on Zhipu AI's GLM-5.3 model, finding that its safety guardrails are weak. The report showed that removing safety filters — a process called abliteration — drops the model's refusal rate for harmful requests to just 6 percent. Running a modified version requires expensive hardware such as Nvidia H200 chips and costs thousands of dollars, which Anthropic says creates a high barrier for ordinary hackers but remains accessible to well-resourced actors.
Meanwhile, a Businessolver study of 300 C-suite leaders found that executives focused on cutting jobs with AI are half as likely to invest in AI upskilling. About 27% of leaders prioritize cost savings through headcount reduction, yet only 18% also prioritize AI training. That compares with 35% of leaders who do not focus on job cuts. The study warns that reducing staff without building skills may undermine productivity gains from AI adoption.
Key Takeaways
- <ul><li>The SEC charged four entities with running a $15 million AI trading scam targeting hundreds of retail investors through fake WhatsApp and Facebook groups</li><li>Cryptoaiml and TSAI entities misappropriated over $12.5 million and $2.8 million respectively, with no real trading occurring behind promised profits</li><li>Robinhood launched a 24/7 AI-powered trading agent and new AI tools for active traders, with shares rising 4.5% pre-market</li><li>Anthropic found that Zhipu AI's GLM-5.3 model drops to a 6% refusal rate on harmful requests when safety filters are removed</li><li>Running a modified GLM-5.3 model requires expensive Nvidia H200 hardware and thousands of dollars per instance</li><li>A Businessolver study shows executives pursuing AI job cuts are 50% less likely to invest in AI training for their teams</li><li>Only 18% of leaders focused on headcount reduction also prioritize AI upskilling, versus 35% of those not cutting jobs</li><li>The SEC urges investors to use Investor.gov to verify the legitimacy of investment providers</li><li>Crypto.com's AI agent platform remains in stealth mode eight months after a $15 million Super Bowl ad launch</li><li>HPE CFO says AI demand continues to outpace supply, with the company being selective on large deals to protect margins</li></ul>
SEC sues Cryptoaiml and TSAI over $15M AI trading fraud
The SEC alleges Cryptoaiml and TSAI took more than $15.3 million from hundreds of retail investors using fake AI trading bots. The entities used WhatsApp groups from August 2024 through March 2025 to build trust and promise large profits. They falsely claimed SEC oversight and posted fake regulatory documents. No real trading occurred and displayed profits were fictitious according to the SEC.
SEC exposes $15 million alleged scam targeting investors
The SEC charged Cryptoaiml Ltd., Cryptoaiml Capital Foundation, TSAI Pro Ltd., and TSAI Capital Foundation on September 29. The agency says the entities misappropriated more than $12.5 million and $2.8 million respectively from hundreds of retail investors. WhatsApp groups and Facebook were used to lure victims with promises of AI-powered profits. The SEC encourages investors to use Investor.gov to verify investment providers.
SEC charges four entities over $15M crypto AI scam
The U.S. Securities and Exchange Commission filed cases in Manhattan on September 29 against four entities. The agency alleges the groups raised more than $15 million using WhatsApp and Facebook to attract investors with promises of high returns. Operators falsely claimed SEC oversight and used fabricated documents. Investors faced frozen accounts and were asked to pay additional fees to withdraw funds.
Robinhood expands AI tools for active traders
Robinhood CEO Vlad Tenev announced new AI-powered trading tools at the HOOD Engines of Creation summit. The suite of tools will be integrated into the Robinhood app to help traders make more informed decisions. This move is part of Robinhood's broader push to become the go-to platform for active traders. The company has been investing heavily in AI capabilities.
Robinhood unveils AI agent and 24/7 trading
Robinhood launched a new AI-powered trading agent that operates 24/7 at its annual summit. The agent provides real-time market data and analysis to help traders make better decisions. Robinhood shares rose 4.5% in pre-market trading following the announcement. The company has been working on the AI agent project for several months.
AI Job-Cutting Executives Less Likely to Invest in AI Training
A Businessolver study of 300 C-suite leaders and 1,000 employees found that executives focused on cutting jobs with AI are half as likely to invest in AI upskilling. About 27% of leaders prioritize cost savings through headcount reduction, but only 18% also prioritize AI training. That compares with 35% of leaders who do not focus on job cuts. The study warns that reducing staff without building skills may hurt productivity gains.
Crypto.com AI Agent Still Not Ready After Super Bowl Launch
Crypto.com's personal AI agent platform on ai.com is still under construction nearly eight months after a Super Bowl ad debut. The $15 million ad in February caused the website to crash as users tried to sign up. A spokesperson confirmed the platform is still in stealth mode. Rivals like Robinhood, Kraken, and Bitget have already launched narrower AI trading tools. Crypto.com has also cut about 12% of its workforce due to AI integration.
VeriWeave Govern Tests Safety Controls for Enterprise AI Agents
VeriWeave Govern is a runtime governance layer that checks AI agent actions against set policies before approval. It uses evidence validation, deny-by-default rules, and human review for important decisions. In tests on 60,000 cases across five fields, it achieved 0.9888 mean accuracy and zero false allows. The system also passed 12 end-to-end scenarios and 40,040 concurrent requests with no failures. Researchers note a safety versus usefulness trade-off compared to human judgment.
SBA Watchdog Criticizes Agency's AI Readiness
The Small Business Administration inspector general found that the SBA still lacks a full AI strategy, a generative AI policy, and an active AI governance board. The agency also has no process to identify high-impact AI uses, including a fraud detection pilot the watchdog classified as high impact. The OIG issued six recommendations, and the SBA agreed to all of them. SBA Deputy CIO Douglas Robertson said the agency is working to meet federal AI requirements while balancing innovation and safety.
HPE CFO Says AI Demand Still Outpaces Supply
Hewlett Packard Enterprise CFO said demand for AI continues to outpace supply and the company remains in a supply constraint environment. HPE is being selective about large deals to protect profit margins. The CFO pointed to growth in operating profit and EPS as proof of this approach. He also said HPE is positioning itself as a growth company with new opportunities in networking. The company expects this to drive future revenue growth.
TOCOMAS creates consistent multi-agent systems
Researchers introduced TOCOMAS, a new system that keeps multi-agent tasks organized and consistent. This tool links task steps, agent roles, and memory rules so they match each other perfectly. During testing on four different benchmarks, TOCOMAS improved success rates compared to older methods. The system also helps agents evolve their own rules while staying within safe structural limits.
Context Language Models manage their own data
Scientists created Context Language Models, or CLMs, which handle their own context like a personal file. These models can update their memory automatically and work well in groups of multiple agents. Tests show CLMs achieve higher accuracy while using significantly less computing power than current top strategies. An online learning method further improved performance on specific tasks while reducing energy use.
Anthropic warns Chinese AI model has weak safety
Anthropic released a report claiming Zhipu AI's GLM-5.3 model has weak safeguards against creating harmful content. The company found that removing safety filters, a process called abliteration, drops the refusal rate for bad requests to just 6 percent. Running this modified model requires expensive hardware like Nvidia H200 chips and costs thousands of dollars. Anthropic argues that while nation-states might afford such setups, ordinary hackers face high costs and technical barriers.
Sources
- SEC Sues Cryptoaiml and TSAI Over Alleged $15M AI Trading Fraud
- SEC exposes $15 million alleged scam: How WhatsApp groups, AI trading promises and fake SEC claims lured i
- SEC Charges Four Entities Over Alleged $15M Crypto, AI Investment Scam
- Robinhood deepens agentic AI offering in push to become go-to place for active traders
- Robinhood Stock Pops After Unveiling AI Agent, 24/7 Trading
- Executives Using AI to Cut Jobs Are Half as Likely to Invest in AI Upskilling, Study Shows
- Crypto.com AI agent still in ‘stealth mode’ nearly 8 months after Super Bowl debut
- VeriWeave Govern: Evidence-Gated Deterministic Runtime Governance for Enterprise AI Agents
- SBA gets another push to step up its AI game
- AI demand is still outpacing supply, HPE CFO explains
- Topological Coherence for Self-evolving Multi-agent Systems
- Context Language Models
- Anthropic claims popular Chinese AI model has Mythos-class hacking abilities
Comments
Please log in to post a comment.