Microsoft's Codename MDASH is drawing real-world attention, with Tanium joining the private preview to test the AI security platform. MDASH uses over 100 specialized AI agents across multiple models to uncover code vulnerabilities like logic flaws and authorization problems that traditional rule-based tools overlook. Tanium ran MDASH against its own production code, pairing AI scanning with human review. Christian Hunt and Loic Simon emphasized that AI works best when combined with the judgment of security engineers rather than replacing it.
Meanwhile, Nvidia appears to be deepening its involvement with Reflection AI, where CEO Misha Laskin confirmed reports that the chipmaker is expanding its stake in the company. Laskin also discussed the release of Beam, an open-weight AI model, and openly acknowledged that the risks of artificial intelligence certainly worry him. The conversation aired on The Claman Countdown on Fox Business.
In the government sector, David Chou, vice president at Maximus, argued that proving AI actually works is harder than deploying it. He urged federal agencies to track metrics like workforce adoption, repeat usage, and service delivery improvements. In contact centers, he said AI is performing well if call quality stays within four points of human performance while handling larger volumes. He also stressed that employees must continue fact-checking AI outputs and recommended data ontologies to help define and map organizational data for clearer AI reasoning.
On the advertising side, a system called AdCoEvo is showing measurable gains. The co-evolving multi-agent system automates advertising delivery by connecting decisions across different timescales through role specialization. Experiments found that AdCoEvo increased advertiser value by 14.02% and reduced cost per acquisition by 7.07% compared to rule-based methods, improving strategies through feedback among planning, adjustment, and diagnosis agents.
Vertafore launched two AI agents for the insurance industry: the Velocity Policy Intake Agent and the Velocity Policy Check Agent. The Intake Agent extracts policy details from documents and loads them into agency management systems, while the Check Agent compares issued policies against existing records to flag inconsistencies. Early trials showed time savings of up to 87% for policy intake and 90% for policy checking. Both agents are part of Vertafore's Agentic Agency strategy and are available in AMS360, with early access for Sagitta.
UPS Capital introduced CommerceShield, an AI model trained on 26 billion delivery touchpoints across 150 million U.S. addresses to predict delivery risks before packages ship. A UPS Capital survey found that 60% of consumers would stop shopping at a retailer after a bad shipping experience, while 61% would trust a retailer more if a missing package were replaced in time. The tool also creates a multi-signal super score using factors like device and VPN data to help retailers take actions such as requiring signatures or routing deliveries to staffed pickup points.
In crypto markets, a survey found 68.9% of traders use AI for research before executing trades, yet trust and security concerns are limiting wider adoption. AI Agent tokens dropped 7% in 24 hours even as trading volume rose 28.4%, showing that traders embrace AI for research but remain cautious about letting it trade independently.
A new study across 21 model-task configurations examined whether latent chain-of-thought in AI actually leads to better reasoning. On logical tasks, the chain often carried an answer already readable before reasoning began, and deleting the chain left predictions almost unchanged. On arithmetic tasks, all tested recipes showed dependence on chain content. Researchers introduced a taxonomy separating what a chain represents from how the model uses it, arguing that readable latent states alone are not sufficient evidence of reasoning.
Credo, a small-cap company selling copper and optical connections to five hyperscalers for AI clusters, is positioned for growth as reliability between processors becomes increasingly critical for AI operations. The company has strong revenue growth, higher margins than peers, and the lowest market cap in its group. The 24/7 Wall St. price target is $247.25 with a buy rating.
Beyond large language models, modern machine learning is reshaping how investors make decisions. These tools analyze large amounts of data to help with portfolio management by finding patterns in historical data and market trends, while also improving risk management by spotting potential dangers early. Machine learning is becoming a key tool for smarter investing across the industry.
Tanium also participated in Microsoft's AI security preview, reinforcing the growing trend of enterprises testing AI-driven vulnerability detection. Tanium provided feedback on MDASH while running it against its own codebase with human review, underscoring that real-world validation remains essential as AI security tools mature.
Key Takeaways
- <ul><li>Tanium joined Microsoft's private preview for Codename MDASH, an AI security system using over 100 specialized AI agents to find code vulnerabilities like logic flaws and authorization problems</li><li>Nvidia is reportedly expanding its stake in Reflection AI, where CEO Misha Laskin released the open-weight model Beam and expressed concern about AI risks</li><li>David Chou at Maximus says federal agencies should track workforce adoption, repeat usage, and service delivery improvements to prove AI tools actually work</li><li>AdCoEvo increased advertiser value by 14.02% and reduced cost per acquisition by 7.07% compared to rule-based advertising methods</li><li>Vertafore's Velocity Policy Intake and Check Agents showed time savings of up to 87% and 90% respectively in early insurance trials</li><li>UPS Capital's CommerceShield AI model is trained on 26 billion delivery touchpoints across 150 million U.S. addresses to predict delivery risks before shipping</li><li>68.9% of crypto traders use AI for research but trust and security concerns limit independent AI trading adoption</li><li>A study across 21 model-task configurations found that latent chain-of-thought does not necessarily prove AI reasoning, as readable latent states alone are insufficient</li><li>Credo, a small-cap company selling connections to five hyperscalers for AI clusters, has a 24/7 Wall St. price target of $247.25 with a buy rating</li><li>Modern machine learning is transforming investor decision-making by analyzing large datasets for portfolio management and early risk detection</li></ul>
Tenium tests Microsoft AI security system MDASH
Tanium joined Microsoft's private preview for Codename MDASH, an AI security system using over 100 specialized AI agents. The system finds code vulnerabilities like logic flaws and authorization problems that traditional tools miss. Tanium tested MDASH against its own production code and combined AI scanning with human review. Christian Hunt and Loic Simon highlighted the importance of combining AI with security engineer judgment.
Tanium participates in Microsoft AI security preview
Tanium became a testing partner for Microsoft's Codename MDASH security platform. MDASH uses more than 100 specialized AI agents across multiple models to find code vulnerabilities. The tool identifies issues like logic flaws and cross-file weaknesses that rule-based methods miss. Tanium provided feedback on the system while running it against its own codebase with human review.
Credo stock poised for AI connection growth
Credo is a small-cap company selling copper and optical connections to five hyperscalers for AI clusters. The company focuses on reliability between processors, a critical issue for AI operations. Its revenue growth is strong, and it has higher margins than peers while holding the lowest market cap. The 24/7 Wall St. price target is $247.25 with a buy rating.
AdCoEvo system improves AI advertising delivery
AdCoEvo is a co-evolving multi-agent system designed to automate advertising delivery operations. The system connects decisions across different timescales using role specialization among agents. Experiments showed AdCoEvo increased advertiser value by 14.02% and reduced cost per acquisition by 7.07% compared to rule-based methods. It improves strategies through feedback among planning, adjustment, and diagnosis agents.
Crypto traders use AI for research but question trust
A survey found 68.9% of crypto traders use AI for research before executing trades. AI Agent tokens dropped 7% in 24 hours even as trading volume rose 28.4%. Trust and security concerns are limiting wider adoption of AI tools in crypto trading. Traders embrace AI for research but remain cautious about letting it trade independently.
UPS Capital launches AI tool to prevent holiday delivery failures
UPS Capital released CommerceShield, an AI model trained on 26 billion delivery touchpoints across 150 million U.S. addresses to predict delivery risks before packages ship. A UPS Capital survey found 60% of consumers would stop shopping at a retailer after a bad shipping experience, and 61% would trust a retailer more if a missing package were replaced in time. Only 10% of consumers viewed shipping protection as a hidden fee, while over half said it signals accountability. The tool also creates a multi-signal super score using factors like device and VPN data to help retailers take actions such as requiring signatures or routing deliveries to staffed pickup points.
Proving AI works is harder than deploying it, expert says
David Chou, vice president at Maximus, said federal agencies and other organizations must focus on measuring whether AI tools actually improve results rather than just launching them. He recommended tracking workforce adoption, repeat usage and service delivery improvements to determine success. In contact centers, he said AI is on track if call quality stays within four points of human performance while handling larger volumes. He also stressed that employees still need to fact-check AI outputs and that data ontologies can help by defining and mapping organizational data for clearer AI reasoning.
Reflection AI CEO discusses AI risks and new open-weight model
Reflection AI CEO Misha Laskin spoke about the release of the open-weight AI model Beam, the company's valuation and reports that Nvidia is expanding its stake in the firm. He expressed concern about the risks of artificial intelligence, saying it certainly worries him. The discussion took place on The Claman Countdown. The video is available on Fox Business.
Study examines whether latent chain-of-thought truly reasons in AI
A new study across 21 model-task configurations examined whether latent chain-of-thought, which replaces written reasoning steps with hidden states, actually leads to better answers. On logical tasks, the chain often carried an answer already readable before reasoning began, and deleting the chain left predictions almost unchanged. On arithmetic tasks, all tested recipes showed dependence on chain content. The researchers introduced a taxonomy separating what a chain represents from how the model uses it, arguing that readable latent states alone are not sufficient evidence of reasoning.
Vertafore launches AI agents for insurance policy checks
Vertafore launched the Velocity Policy Intake Agent and Velocity Policy Check Agent to automate insurance policy data entry and verification. Early trials showed time savings of up to 87% for policy intake and 90% for policy checking. The Intake Agent extracts policy details from documents and loads them into an agency management system, while the Check Agent compares issued policies against existing records to flag inconsistencies. Both agents are part of Vertafore's Agentic Agency strategy and are available in AMS360, with early access for Sagitta.
Modern Machine Learning Is Transforming How Investors Make Decisions
Modern machine learning is changing the investing world beyond just large language models. These tools can quickly analyze large amounts of data to help investors make better choices. Machine learning helps in portfolio management by finding patterns in historical data and market trends. It also improves risk management by spotting potential dangers early. Overall, machine learning is becoming a key tool for smarter investing.
Sources
- Tanium joins Microsoft AI security preview for MDASH
- Tanium joins Microsoft AI security preview for MDASH
- One Stock That Could Become Much More Important to AI
- AdCoEvo: A Co-Evolving Multi-Agent Advertising Delivery System
- Crypto traders embrace AI for research – But can they trust it to trade?
- UPS Capital Uses AI to Stop Holiday Delivery Disasters
- Protected: The hard part of AI isn’t deployment
- Reflection AI CEO confronts the risks of artificial intelligence: ‘It certainly worries me’
- Does Latent Chain-of-Thought Actually Reason? A Single, Unifying Taxonomy for Latent Reasoning
- Vertafore launches AI agents to automate insurance policy checks
- Beyond LLMs: Modern Machine Learning is Changing Investing
Comments
Please log in to post a comment.