A lawsuit over an AI gun-detection alert at Kenwood High School in Baltimore County says Omnilert’s system mistook a student’s bag of Doritos for a weapon in 2025. Four current and former students say police approached them with guns drawn, made them kneel, handcuffed them and searched one student. Filed Nov. 17 in Baltimore County Circuit Court, the case alleges unreasonable searches and seizures, false arrest and imprisonment, and negligence. The students seek unspecified damages and an order barring the district from using Omnilert until it improves training and safety procedures. The alert was canceled after being sent for human review; Omnilert says the system operated as designed. School officials and police declined to comment on the pending case.
School districts are also setting broader rules for AI use. Sarasota’s school board adopted an AI policy by a 3-to-1 vote despite students’ objections that it was vague and contradictory. In Kansas, USD 435 established guidelines intended to help teachers and students use AI safely and effectively. UC Berkeley, meanwhile, plans to pilot a machine-learning grading tool this spring with as many as 15 Haas School of Business faculty members. Professors will check its grading of exams, quizzes and assignments in courses including statistics and finance; the campus may expand the pilot if it succeeds.
For physical security teams, Lumana and HiveWatch have linked AI video alerts to HiveWatch’s Global Security Operations Center platform. Lumana analyzes feeds from existing IP cameras and sends detected events with video to a shared workspace, where operators can follow response procedures. The companies say the integration could reduce alert fatigue and speed response, but have not provided measured results or pricing. Guidance for corporate security buyers emphasizes testing threat detection, alert explanations, source and timestamp visibility, workflow routing, escalation and delivery failures using realistic events such as fires, violence and road closures.
A Harvard study estimates that the 400 largest U.S. data centers emitted 28.3 million tons—25.7 million metric tons—of carbon dioxide in one year, exceeding Denmark’s annual emissions and coming in two to three times above earlier estimates. More than half of their electricity came from fossil fuels. At U.N. climate talks, Turkey and the International Telecommunication Union launched the Antalya Artificial Intelligence Pledge, calling for clean power, efficient energy and water systems, and responsible AI. Al Gore argued data centers need carbon-free power while noting AI could help cut emissions through efficiency.
Investors are weighing AI’s growth against market and reporting risks. About 83% of analyst recommendations for major AI-related companies are Strong Buy and 2% are Sell, a pattern the market-crash analysis compares with the optimism before the dot-com collapse and the 2008 financial crisis. The article cautions that demand for strong technology does not by itself justify high stock prices, while the timing and scale of any downturn remain uncertain. Separately, Anthropic reported $33.2 billion in revenue for the first nine months of 2026 in an amended filing, compared with $16.3 billion in the first half. Usage-based sales—measured in tokens—made up about 80% of first-half revenue. Anthropic says more business customers have multiyear minimum-spend contracts, but it recognizes revenue as tokens are used. Without token volumes or revenue per token, investors cannot distinguish growth in actual usage from unused commitments.
Microsoft CEO Satya Nadella said a software bug exposed by a particular traffic pattern caused the previous week’s global Outlook outage, and that testing should have caught it. He said AI helps teams find and fix bugs, though patches can introduce instability, especially when applied to older infrastructure. Separately, JPMorgan, General Motors and Pfizer joined a letter to the White House seeking more business involvement in AI policy, arguing that rules affect the wider economy and future of work, not only technology companies.
Research and infrastructure developments round out the news. An open-access journal published by the American Institute of Artificial Intelligence is seeking work for a March 2027 special issue on AI and software supply-chain security, including vulnerability detection, code verification and component tracing. LanceDB announced $57.5 million in Series B funding on Nov. 17, 2026, bringing its total funding to $70 million. It plans to expand its multimodal lakehouse and tools for managing training data, feature engineering, experiments and data loading. The company says its platform handles video, images and sensor data, supports more than 130,000 requests per second and training across 8,000 GPUs, and has seen more than tenfold revenue growth since June 2025. Its users and customers include Netflix, Microsoft AI, Runway and NVIDIA Cosmos.
Key Takeaways
- Four Baltimore County students sued over an Omnilert AI alert that allegedly mistook a Doritos bag for a gun and led police to detain and search them; the Nov. 17 lawsuit seeks damages and changes to the district’s use of the system.
- Sarasota adopted a district AI policy 3–1 despite student criticism, while USD 435 established classroom-use guidelines.
- UC Berkeley will test an AI grading tool with up to 15 Haas faculty members this spring, with professors checking its results.
- Lumana and HiveWatch integrated AI video alerts with HiveWatch’s security operations platform; no measured performance results or pricing were disclosed.
- A Harvard study estimates the 400 largest U.S. data centers emit 25.7 million metric tons of COâ‚‚ annually, with more than half their electricity from fossil fuels.
- Anthropic reported $33.2 billion in revenue for the first nine months of 2026; usage-based sales represented about 80% of first-half revenue, but token-volume and revenue-per-token data are absent.
- About 83% of analyst recommendations for major AI-related companies are Strong Buy, prompting comparisons with market optimism before earlier crashes; the timing and scale of a possible downturn are uncertain.
- Microsoft CEO Satya Nadella said testing should have caught the bug behind the global Outlook outage and noted that patches can create instability in older infrastructure.
- JPMorgan, General Motors and Pfizer signed a letter asking the White House for greater business influence in AI policy.
- LanceDB raised $57.5 million in Series B funding, bringing total funding to $70 million; customers and users include Microsoft AI, Runway and NVIDIA Cosmos.
Lawsuit says AI chip alert led to armed police encounter at school
Four teens are suing Baltimore County schools and police after Omnilert’s AI system mistook a bag of Doritos for a gun at Kenwood High School. The alert was canceled within minutes, but the lawsuit says Principal Catherine Smith later sent a video image to a school resource officer, leading to an armed police response. Officers surrounded the teens, forced them to the ground, handcuffed and searched them. The complaint, filed in Baltimore County Circuit Court, alleges violations of the teens’ rights and says the school failed to verify the alert. Omnilert, police and school officials declined to comment on the pending case.
Sarasota adopts AI policy despite students’ concerns
Sarasota’s school board adopted a policy on how the district will use artificial intelligence in schools by a 3 to 1 vote. Several high school students spoke against the plan, saying it had flaws and contradictions and was too vague. The vote took place at the board’s first meeting with its newly elected members. Tom Edwards was named chair and Liz Barker was named vice chair, while Edwards was absent from the meeting.
Students sue Baltimore County schools over AI gun detection error
Students at a Baltimore County high school are suing the school system after an AI detection system mistook a student’s bag of chips for a gun. Their lawsuit says the false alert led to a school lockdown and violated their Fourth Amendment rights against unreasonable searches and seizures. The students’ parents say the system is flawed and are seeking damages and changes to the district’s rules for using AI detection systems. The school system said the technology is designed to identify possible threats and that the students were not targeted. The lawsuit is ongoing.
Baltimore County students sue over AI system’s Doritos false alarm
Current and former Baltimore County students are suing the school district, police and Omnilert over an AI system that mistook a bag of Doritos for a gun at Kenwood High School in 2025. The lawsuit says police approached four teenagers with guns drawn, detained them and searched one student. The students accuse the defendants of unreasonable searches and seizures, false arrest, false imprisonment and negligence, and seek unspecified damages. They also want a court order stopping the district from using Omnilert until it improves training and safety procedures. Omnilert said its system sent the alert for human review, then marked the event resolved, and operated as designed.
Maryland students sue after AI mistakes Doritos bag for a gun
Students have sued after Baltimore County police handcuffed and searched Taki Allen and his friends outside Kenwood High School in Essex, Maryland. Allen said officers made him kneel and put his hands behind his back before they found he had no weapon. The lawsuit, filed Nov. 17 in Baltimore County Circuit Court, says malfunctioning technology and the actions of school administrators and police violated the students’ constitutional rights. Baltimore County Public Schools began using Omnilert’s AI security system in 2023, and the system uses existing school cameras to detect weapons. Omnilert said the process worked as intended by sending the alert for rapid human verification.
USD 435 sets rules for AI use in classrooms
USD 435 has established guidelines for teachers and students using artificial intelligence in education. The district aims to help them use the technology safely and effectively as AI becomes more common in schools.
UC Berkeley tests AI exam grading with up to 15 Haas professors
UC Berkeley will pilot an AI grading tool this spring semester with up to 15 faculty members at the Haas School of Business. The tool uses machine learning to grade exams, quizzes and assignments, with professors checking its accuracy. It will be tested in several classes, including statistics and finance. The campus has not shared the pilot’s results and may expand the tool if it succeeds.
Lumana and HiveWatch link AI video alerts to security teams
Lumana and HiveWatch have completed an integration that sends AI-generated video alerts into a shared workspace for physical security teams. Lumana analyzes video from existing IP cameras and sends detected events, with video footage attached, to HiveWatch’s Global Security Operations Center platform. HiveWatch connects information from security systems and guides operators through response procedures. The companies say the link may reduce alert fatigue and speed up responses, but they did not provide measured results or pricing. The setup requires configuring an HTTP integration and connecting cameras and analytics alerts.
How to assess AI platforms for corporate security
Corporate security teams at facilities such as semiconductor plants and robotics sites should assess how AI platforms detect threats, explain alerts and fit existing workflows. The article recommends testing real-time detection of nearby violence, road closures and fires, including events on employee routes. Buyers should check whether systems connect supporting reports, show sources and timestamps, and keep uncertainty visible for analyst review. Demonstrations should also show how alerts reach the right security manager, how they are acknowledged or escalated, and what happens if delivery fails. The article says the right platform depends on the operational need and includes external event intelligence and physical security management options.
US data centers emit more pollution than Denmark, study estimates
A Harvard University study estimates that the 400 largest U.S. data centers produced 28.3 million tons, or 25.7 million metric tons, of carbon dioxide in one year. That is more than Denmark’s yearly emissions, and the estimate is two to three times higher than earlier studies. More than half of the centers’ electricity came from fossil fuels, and senior author Francesca Dominici warned that plans for 2,000 more centers could increase the concern. At U.N. climate talks, Turkey and the International Telecommunication Union launched the Antalya Artificial Intelligence Pledge, which calls for clean power, efficient energy and water systems, and responsible AI. Former U.S. Vice President Al Gore said data centers need carbon-free power, while expressing hope that AI could help reduce emissions by improving efficiency.
AI Stock Optimism Faces Warnings From Past Market Crashes
Wall Street remains strongly optimistic about artificial intelligence stocks, even as concerns grow about profits, borrowing costs and company risks. About 83% of analyst recommendations for major AI related companies are Strong Buy, while 2% are Sell. The article compares this outlook with the dotcom bubble, when roughly 85% of analyst recommendations for major technology companies were Strong Buy before the market crash, and with the 2008 financial crisis, when bank stocks still received mostly favorable ratings despite warning signs. It says AI companies may benefit from lasting demand, but strong technology does not automatically make high stock prices reasonable, and the timing and scale of any downturn remain uncertain.
Why Token Use Could Matter to Investors in an Anthropic IPO
Most of Anthropic’s revenue comes from customers paying for how much they use its AI models, measured in tokens. In an amended filing, the company reported $33.2 billion in revenue for the first nine months of 2026, up from $16.3 billion in the first half of the year; usage based sales made up about 80% of revenue in that first half. Anthropic says more business customers have signed multiyear contracts with minimum spending commitments, but it records revenue only as customers use tokens. Without total token volumes or revenue per token, investors cannot assess how much growth reflects actual demand or unused commitments. Bondcoins founder Claire Husson-Citanna says clearer reporting could help investors judge future revenue and costs and may lower AI labs’ borrowing costs.
Satya Nadella Says Testing Could Have Prevented Outlook Outage
Microsoft CEO Satya Nadella said a software bug caused the previous week’s Outlook outage when a particular traffic pattern exposed it. In an interview with CNBC’s Jim Cramer, he said Microsoft should have found the bug through testing and prevented a global outage. Nadella also said AI helps teams find and fix bugs, but the required patches can create instability. He described some instability as the result of applying new patches to older infrastructure.
JPMorgan, GM and Pfizer Seek More Influence Over AI Policy
JPMorgan, General Motors and Pfizer are among the companies that signed a letter to the White House asking for greater involvement in artificial intelligence policy. The companies argue that businesses using AI should have more say in how it is developed and regulated. The letter says AI policy affects not only the technology industry but also the future of work and the economy. JPMorgan, GM and Pfizer are described as the largest US bank, automaker and pharmaceutical company, respectively.
AI Journal Seeks Research on Software Supply Chain Security
A peer reviewed, open access journal published by the American Institute of Artificial Intelligence is seeking submissions for a March 2027 special issue on software supply chain security. The call covers using AI to find vulnerabilities and methods to verify code as software is built and released. It also welcomes research on tracing software components and assessing risks linked to AI models and the software they depend on. The journal seeks original research, technical studies, policy analyses and practical case studies relevant to the public sector or national security. Authors should identify the special issue when submitting and can contact the editorial office for the submission timetable.
LanceDB raises $57.5 million to expand its AI data platform
LanceDB announced on November 17, 2026, that it raised $57.5 million in Series B funding, bringing its total funding to $70 million. It plans to expand its multimodal lakehouse, which organizes and delivers AI training data, and develop tools for training-data management, feature engineering, experiments, and data loading. The platform handles videos, images, and sensor data, and LanceDB says it supports more than 130,000 requests per second, tables with hundreds of petabytes or more than 40 billion vectors, and training across 8,000 GPUs. The company reported more than tenfold revenue growth since its Series A in June 2025, and more than 120 million downloads of its open-source library. Customers and users include Netflix, Microsoft AI, Runway, and NVIDIA Cosmos.
Sources
- Baltimore County schools, police sued for AI software 'bag of chips' mistake
- Sarasota students warn of harms of AI in schools, as district adopts new policy
- Students sue Baltimore County Schools after AI detection system mistook chips for gun
- Baltimore County students sue over AI system that mistook Doritos bag for gun
- Maryland students sue after AI security mistook bag of Doritos for gun
- USD 435 develops guidelines for artificial intelligence use in classrooms
- AI grading rolls out at UC Berkeley as Haas pilots automated exam-grading tool with up to 15 professors
- Lumana and HiveWatch Connect AI Video Alerts With Security Operations
- What to Look for in an AI-Powered Corporate Security Platform
- Study estimates magnitude of US data center pollution as climate talks unveil clean AI pledge
- Is the AI Stock Boom Heading for Another Dotcom-Style Crash?
- Anthropic IPO: Why token usage could become a key metric for AI investors
- Satya Nadella admits Microsoft could have prevented Outlook outage: Is AI making software less stable?
- Exclusive: JPMorgan, GM, Pfizer want bigger say in AI policy
- AI journal calls for papers on software supply chain security
- LanceDB Raises $57.5M Series B to Expand Its Multimodal AI Lakehouse
Comments
Please log in to post a comment.