Commvault and Zscaler are joining forces to help companies discover AI use, oversee AI activity and recover from harmful actions by AI agents. The partnership connects Zscaler’s Security for AI and Zero Trust cloud with Commvault tools for restoring data, applications and identities, including automated rollback. Zscaler says its Security for AI pipeline grew more than 75% quarter over quarter, with most deals also including Data Security. Its shares have shown mixed movement: down 12.9% over the past month and 16% year to date in one report, but up 5.9% over 90 days and 5.3% ahead of its December 2, 2026 earnings report. Investors are watching for evidence of demand as slower spending and execution challenges remain concerns.
Google’s Threat Intelligence Group reports that attackers are using AI to identify software vulnerabilities, develop exploits and launch attacks with limited or no human involvement. It found exploit_agent on an exposed staging server: the tool used large language models to compare software before and after updates, then search for weaknesses. The server also contained reconnaissance, password-guessing and webshell scripts, along with stolen data suggesting interest in French media, entertainment and education sites. Google did not establish that a review of a WordPress plugin resulted in an attack. It also reported AI use in attacks on cryptocurrency companies by North Korean group MIDNIGHT NEPTUNE. Google says such tools can leave detectable traces, while warning defenders they may need AI agents to keep pace.
Identity security and AI infrastructure are also drawing attention. Semperis launched Lightning, a platform covering human, non-human and AI identities across Active Directory, Microsoft Entra ID and Okta. It combines identity inventories, exposure checks, attack-path analysis and detection, with more than 300 security indicators and a live identity graph. Separately, a University of Cambridge study mapped 14 AI supply-chain layers and 137 elements across more than 50 countries. Reviewing 117 policy documents from 43 countries and territories, the EU and OECD, researchers found that policies often emphasize models and data centers while overlooking factories, chemicals and chip-design software. Their toolkit is intended to help governments assess dependencies and consider supplier diversification or local capacity.
Several developments focus on how organizations and users put AI to work. Hitachi Vantara is building industry-specific solutions on its Hitachi iQ stack, with storage tools to classify, govern and serve data. Its object store can hold CT scan metadata in Apache Iceberg format so applications can query scans directly, without a separate ETL process or database. A Black Friday offer, meanwhile, prices 1min.AI’s 20-seat Advanced Business Plan lifetime subscription at $64.97 through December 6 at 11:59 p.m. PT. The app combines models from OpenAI, Anthropic, Google, Meta and Mistral for writing, research, image, audio and video tasks; the plan includes four million monthly credits plus up to 450,000 extra credits for daily logins.
AI agents are moving into other settings, from trading to payments. A London Summit panel on November 24 will discuss the shift from trading copilots to more autonomous colleagues. At least ten firms connected AI agents to live trading accounts in the first half of 2026, often through Model Context Protocol; seven of 14 brokers tested by StockBrokers.com now run their own MCP server. MetaTrader 5 has added native AI agents and MCP support in its July beta. Coinbase CEO Brian Armstrong said the company is building financial accounts and payment rails for AI agents and predicted agents will eventually make 99.9% of transactions. He also argued that teams of open-weight models could counter dominance by a small number of frontier labs.
Research and commercial results round out the developments. A new visual-agent audit method holds public evidence and region bindings fixed while varying observation budgets or task settings, separating errors in what an agent chooses to inspect from errors in its final answer. Researchers tested it across 40 paired routing anchors and five open-weight models. Chinese self-driving startup Pony.ai reported nearly $18 million in robotaxi revenue in the third quarter, prompting questions about the fleet scale Tesla may need; the available report provides no further figures or comparison.
Key Takeaways
- Commvault and Zscaler link AI access controls with data, application and identity recovery; Zscaler says its Security for AI pipeline grew more than 75% quarter over quarter.
- Zscaler’s share performance is mixed, and investors are watching its December 2, 2026 earnings for signs of AI security demand.
- Google found exploit_agent, an AI-enabled tool that can compare software updates, identify vulnerabilities and create or run exploits; Google also reported AI use by North Korean group MIDNIGHT NEPTUNE in attacks on cryptocurrency companies.
- Semperis Lightning covers human, non-human and AI identities across Active Directory, Microsoft Entra ID and Okta, with more than 300 security indicators.
- A Cambridge study found AI supply-chain policies often overlook factories, chemicals and chip-design software, despite mapping 14 layers and 137 elements across more than 50 countries.
- Hitachi Vantara says its storage can serve governed data directly; its object store supports querying CT scan metadata in Apache Iceberg without a separate ETL process or database.
- 1min.AI’s 20-seat lifetime business plan is offered for $64.97 through December 6 and combines models from OpenAI, Anthropic, Google, Meta and Mistral.
- Trading firms are connecting AI agents to live accounts: at least ten did so in the first half of 2026, and seven of 14 brokers tested by StockBrokers.com run their own MCP server.
- Coinbase CEO Brian Armstrong says the company is building accounts and payment rails for AI agents and predicts agents could make 99.9% of transactions.
- Pony.ai reported nearly $18 million in robotaxi revenue in the third quarter; a separate visual-agent audit was tested on 40 paired routing anchors and five open-weight models.
Zscaler Shares Fall as Commvault Partnership Targets AI Security
Commvault partnered with Zscaler to help companies secure their use of AI. The deal combines Zscaler’s Security for AI controls with Commvault’s data recovery tools to support AI discovery, policy oversight, and the rollback of harmful activity by AI agents. Zscaler shares fell 12.9% over the past month and 16% year to date, while its one-year total shareholder return dropped 36.5%. Its share price gained 5.9% over the past 90 days.
Commvault Deal Adds to Zscaler’s AI Security Push Ahead of Earnings
Commvault and Zscaler announced a partnership to secure enterprise AI use by combining AI access controls with tools for recovering data, applications, and identities. Their joint work covers AI discovery, runtime oversight, detecting AI agents, and automated recovery. Zscaler’s Security for AI pipeline grew more than 75% quarter over quarter, and most of those deals included Data Security. Investors will watch Zscaler’s earnings on December 2, 2026, for signs of AI security demand, while the article notes that slower spending and execution challenges could weigh on the company.
Zscaler Shares Rise 5.3% Ahead of Earnings
Zscaler shares rose 5.3% as investors looked ahead to the company’s December 2 fiscal first-quarter earnings report. The article links the move to improving investor sentiment, a favorable recent analyst view, and continued interest in Zscaler’s AI security strategy. A recent partnership focused on AI resilience may also support the view that the company is expanding its platform.
Zscaler and Commvault Team Up on AI Security and Recovery
Commvault partnered with Zscaler to connect Zscaler Security for AI controls with Commvault’s recovery tools. The partnership targets AI discovery, runtime oversight, and fast rollback of harmful AI agent activity for enterprise customers. It links Zscaler’s Zero Trust cloud to AI risk management and data recovery, extending its role beyond user access security. The article says investors will watch how the deal affects Zscaler’s investment outlook.
Google Finds AI System Automating Cyberattacks
Google’s Threat Intelligence Group says attackers have begun using AI systems to find software vulnerabilities, create exploits, and launch attacks without human help. Google found a framework called exploit_agent on an exposed staging server. It uses large language models to examine web applications and compare software before and after security updates to identify and exploit weaknesses. The server also held stolen victim data and scripts for actions such as reconnaissance and installing webshells, and victim data suggested interest in French media, entertainment, and education websites. Google said these tools can leave signs that help defenders detect them, but warned that security teams may need their own AI agents to keep up.
Google uncovers automated AI hacking tool with Russian comments
Google says it found a tool that used large language models to find software flaws and create and run exploits without human help. The tool watched for software updates, compared code before and after a fix, then used the changes to identify weaknesses. It also had scripts for mapping networks, guessing passwords and placing hidden control panels on websites. Google found data suggesting the operator targeted French media, entertainment and education websites, but the report did not say whether the tool’s review of a WordPress plugin led to an attack. Google also reported AI use in attacks on cryptocurrency companies by the North Korean group MIDNIGHT NEPTUNE.
Semperis launches Lightning platform for AI identity security
Semperis launched its Lightning Platform to help organisations protect and recover identity systems, including identities used by AI agents. The platform combines identity inventories, exposure checks, attack path analysis and detection tools for Active Directory, Microsoft Entra ID and Okta. It includes more than 300 security indicators and a live identity graph that maps routes to privileged accounts and other critical assets. Lightning can also flag suspicious Active Directory and Entra ID activity, including Kerberos and NTLM logons. Semperis says the system gives security teams a clearer view of human, non-human and AI identities as businesses adopt more AI tools.
Cambridge study finds major gaps in oversight of AI supply chains
A University of Cambridge study found that governments overlook about a third of the global AI supply chain in their policies. Researchers mapped 14 layers and 137 elements across more than 50 countries, including materials, chip production, servers and trained models. They reviewed 117 policy documents from 43 countries and territories, the EU and OECD, and found that policies often focus on models and data centers while missing factories, chemicals and chip-design software. The report says Japan, Germany and other middle powers hold important supply chain positions that their policies do not fully recognize. Researchers also created a toolkit to help governments assess their dependencies and decide when to diversify suppliers or build local capacity.
1min.AI 20-seat business plan falls to $64.97 for Black Friday
A Black Friday offer cuts the price of 1min.AI’s 20-seat Advanced Business Plan lifetime subscription to $64.97. The service brings AI models from OpenAI, Anthropic, Google, Meta and Mistral into one app, with tools for writing, research, images, audio and video. The plan includes four million credits each month, plus up to 450,000 extra credits monthly for daily logins. Teams can share the plan among up to 20 members and use it in a browser or the Mac, iPhone and iPad apps. The offer ends December 6 at 11:59 p.m. PT.
London Summit panel to examine what comes after AI trading assistants
Brokers spent much of 2026 connecting AI assistants to trading accounts, as retail traders used plain English prompts to place orders and draft strategies. At least ten firms connected AI agents to live accounts in the first half of the year, often using Model Context Protocol, while seven of 14 brokers tested by StockBrokers.com now run their own MCP server. Platform companies have also added AI features, including MetaTrader 5’s native AI agents and MCP support in its July beta. A panel called “AI in Trading 2027: From Copilots to Colleagues” will discuss the changes on November 24 at the London Summit in Old Billingsgate. The panel will consider how AI may change brokers, why some projects succeed or stall, and how firms will train future trading desk leaders.
A new audit separates visual agents’ search and decision errors
Researchers introduce budget paired counterfactual audits to evaluate visual agents that read dense documents. The method keeps public evidence and region bindings fixed while changing the remaining observation budget or declared task settings. It measures the exact value of each reading history and separates losses from choosing what to inspect from losses in the final answer. The article says the approach was tested across forty paired routing anchors and five open weight models.
Pony.ai’s robo-taxi revenue highlights the scale Tesla may need
Chinese self driving start up Pony.ai reported almost $18 million in robo taxi revenue in the third quarter. The article connects Pony.ai’s results to the size of the robo taxi fleet Tesla may need. The supplied article text does not include further details about the results or fleet comparisons.
Hitachi Vantara uses smarter storage to move AI into production
Hitachi Vantara says scattered data, access controls, governance and changing regulations can slow AI projects as they move from pilot to production. The company is building industry specific solutions on its Hitachi iQ stack and adding tools that classify, govern and serve data from storage. Its object store can hold CT scan metadata in Apache Iceberg format, allowing applications to query scans directly without a separate extract, transform and load process or database. Hitachi Vantara’s Raman said storage can provide more value when it handles data intelligence natively instead of moving data elsewhere for tasks such as indexing.
Coinbase CEO Brian Armstrong discusses AI agents and Bitcoin
Coinbase CEO Brian Armstrong spoke with Aoyon Ashraf at the TomorrowX Summit. He explained why Coinbase is building financial accounts and payment rails for AI agents. Armstrong believes agents will eventually make 99.9% of all transactions, rather than humans. He also said open weight models working in teams could counter a future dominated by a small number of powerful frontier labs, and discussed Bitcoin’s bottom and Coinbase’s next act.
Sources
- Why Did Zscaler (ZS) Stock Drop After Its Latest Update?
- Zscaler Stock Faces AI Security Test After Commvault Partnership
- Zscaler Rallies as Investors Position for Earnings and AI Security Momentum
- Zscaler Stock In Focus After AI Security Partnership With Commvault
- Google Warns 'Era Of Autonomous AI Attacks Has Arrived'
- Google finds self-running AI hacking tool with Russian comments in its files
- Semperis launches Lightning to bolster AI identity security
- A third of the global AI supply chain is 'invisible' to governments, study finds
- 1min.AI’s 20-seat business plan drops to $64.97 for Black Friday
- Brokers Spent 2026 Plugging Into AI Assistants. A London Summit Panel Asks What Comes Next
- Choosing What to Observe: Budget-Paired Counterfactual Audits for Credible Evaluation of Visual Agents
- Pony AI Results Show Just How Big Tesla’s Robo-Taxi Fleet Needs to Get
- Apache Iceberg and intelligent storage help AI reach production
- Brian Armstrong on bitcoin’s bottom, AI trading agents and Coinbase’s next act
Comments
Please log in to post a comment.