Edition, Wed, 14 Oct 2026.

FBI arrests OpenAI engineer over a hidden recording device

A 35-year-old Saudi engineer faces a federal charge over a "sophisticated recording device" found in an executive conference room, in what the FBI calls one of its highest counterintelligence priorities.

A 35-year-old Saudi engineer faces a federal charge over a "sophisticated recording device" found in an executive conference room, in what the FBI calls one of its highest counterintelligence priorities.

Abdulrahman Alfozan, who joined OpenAI in July 2025, was arrested and charged with using or endeavoring to use a device to intercept oral communications. The device, covered in toilet paper and installed around September 28, was found plugged into a TV's USB port in an executive meeting room at OpenAI's San Francisco office, where "highly confidential technical, commercial, and strategic information" is discussed. OpenAI says it used its own models to identify the device and the employee who planted it, and CCTV footage allegedly showed Alfozan near the room on multiple dates, at one point looking behind the TV.

OpenAI fired the employee and referred the matter to the FBI. Director Kash Patel said the scheme was to "capture sensitive communications about the company's super intelligence technology" and called protecting American superintelligence innovation "one of our highest counterintelligence priorities." The device reportedly had advanced encryption, audio monitoring, and a data-transfer mechanism designed to evade monitoring; if convicted, Alfozan faces up to five years in prison, and OpenAI says there is no indication he inappropriately accessed or exfiltrated model weights, source code, or customer data.

Google rolls out Gemini 4 Argon, its new frontier model

The release, which Google says ranks #1 on the Vals Index, targets "deep work" across coding, finance, legal, cybersecurity, and science.

The release, which Google says ranks #1 on the Vals Index, targets "deep work" across coding, finance, legal, cybersecurity, and science.

Gemini 4 Argon is now available via API in AI Studio and on the Gemini Platform, and to Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search. Google claims the model leads the Vals Index, a benchmark measuring performance across critical economic sectors, and offers introductory pricing of $2 per million input tokens and $10 per million output tokens, with an industry-leading 1M unified input-output token limit via the API.

Google highlighted internal wins: Argon untangled a ~6M-line legacy folder in Chromium's 40M-line codebase in weeks instead of months, cut in-memory session prep time by 40% in long sessions of the open-source Agent Development Kit, and sped up Gemma 4 E2B inference 2.85x on an A100 via vLLM, from roughly 37,900 to over 108,000 tokens per second. In a companion safety post, Google reports Argon solved 12 of 12 end-to-end simulated cyberattack scenarios (up from 5 in the previous generation), with new real-time misalignment monitoring built on activation probes, and notes Wiz used the model to find and patch security gaps in hospital and government infrastructure.

Anthropic is retiring Git in favor of source control built for AI

With merged code changes jumping from about 2,500 a day in May to over 30,000 a day by September, the company built Scion, a version-control system designed for AI-scale change.

With merged code changes jumping from about 2,500 a day in May to over 30,000 a day by September, the company built Scion, a version-control system designed for AI-scale change.

Anthropic describes Scion as a reinvention of centralized source control, along the lines of Subversion. The company said the history of its main branch alone is nearly 100 GB, a compressed working copy is over 25 GB, and it runs on the order of 25 million code checkouts per day for CI, before counting developer environments, agent sandboxes, and research jobs. A team of seven built the system and completed the migration from Git in five months.

Scion ships alongside "Antique," a new code review system, and Anthropic says it makes extensive use of adversarial review, with a model hunting for flaws while skilled engineers steer Claude away from over-engineering. Asked about the lack of offline support, a Scion engineer replied that since Claude is by far the heaviest user of the system and is necessarily online, offline support just wasn't a priority.

Shopify opens a billion listings to any AI agent

Shoppers can now connect their own AI agents to Shop and let them browse and buy across Shopify's catalog, with each purchase still approved by a human.

Shoppers can now connect their own AI agents to Shop and let them browse and buy across Shopify's catalog, with each purchase still approved by a human.

Shopify announced Tuesday that shoppers can configure their AI agents as "personal shoppers," giving agents access to over a billion listings on and off Shopify storefronts. The agent connects through Shop, Shopify's platform that consolidates orders, tracks packages, and recommends products, and buyers can set rules such as budget guardrails or fast shipping only, with checkout happening inside the conversation via one-click Shop Pay.

The move extends Shopify's September partnership with Meta's Muse agent for agentic checkout to any agent, without a partnership integration, as e-commerce becomes one of the battlegrounds of the AI era. Google and Amazon are both building similar agent-shopping tools, and the feature lands amid an ongoing fight over whether agents should be able to shop on a customer's behalf.

US tech giants warn Europe that ditching them could cost €115 billion

A US industry lobby group including Amazon, Google, and Meta is funding dire warnings over Europe's cloud sovereignty plans.

A US industry lobby group including Amazon, Google, and Meta is funding dire warnings over Europe's cloud sovereignty plans.

Amazon, Google, Meta, and other US tech giants are warning European policymakers that moving workloads away from US hyperscalers could cost up to €115 billion, as EU governments weigh cloud sovereignty options. The Register reports the warning is being funded by a lobby group of US tech companies, making the case that the economics of AI-era compute run strongly in favor of the established hyperscalers.

ServiceNow shuts the door on Decagon

After reports that it was weighing a bid for the AI customer-service agent startup, ServiceNow denied any interest, reaffirming an organic-growth-first stance.

After reports that it was weighing a bid for the AI customer-service agent startup, ServiceNow denied any interest, reaffirming an organic-growth-first stance.

ServiceNow responded to media reports suggesting it had shown interest in buying Decagon by saying flatly that it has "no interest or intention of purchasing" the company. In an emailed statement, spokesperson Amit Zavery said the firm is aware of "misinformation in the market today about our acquisition strategy," and that its approach to M&A "has not changed: we remain an organic growth company first." He added that M&A is "a tool to accelerate our roadmap or open meaningful new markets, not a substitute for innovation," noting the company does not see major strategic gaps that require large-scale acquisitions right now.

The denial lands as investors weigh ServiceNow's recent run of multibillion-dollar deals. Decagon, a customer-service AI startup whose CEO is Jesse Zhang, had crossed $100 million in annualized revenue three years after launch and added more than 100 new global enterprise customers in its previous fiscal year. About 90% of its workloads reportedly run on fine-tuned open-source models rather than frontier models, which the company says are faster and cheaper for narrow, repeatable support tasks.

Analyst flags ServiceNow as a top software pick

With the core business stabilizing, one analyst sees NOW revenue growth accelerating into fiscal 2027 as AI, security, and usage-based products grow.

With the core business stabilizing, one analyst sees NOW revenue growth accelerating into fiscal 2027 as AI, security, and usage-based products grow.

A recent analyst note named ServiceNow a top software pick, arguing revenue growth could accelerate into fiscal 2027 as the core business steadies and AI, security, and usage-based offerings become larger contributors to the top line. The framing leans on the view that NOW's platform momentum is broadening beyond its traditional IT service management base, with usage-based pricing opening a larger revenue pool.

Software stocks rally; ServiceNow up 3% on BofA's Salesforce call

ServiceNow climbed alongside a broader software bid as Bank of America flipped Salesforce to Buy and called the AI-disruption worry overblown.

ServiceNow climbed alongside a broader software bid as Bank of America flipped Salesforce to Buy and called the AI-disruption worry overblown.

ServiceNow rose 3% to $144.39 in morning trading, trailing the gain in Salesforce, which jumped after Bank of America double-upgraded the shares to Buy from Underperform and lifted its price target to $285. The bank's argument is that large enterprises are unlikely to replace the platforms housing their business logic, data, and workflows, which counters the fear that AI would disintermediate enterprise software. Palantir lagged, down 0.2% to $208.60, and the iShares Expanded Tech-Software ETF (IGV) added 0.6%.

In the same roundup, ServiceNow's fundamentals were set against its peers: Q2 FY26 revenue grew 24% to $3.99 billion, its AI business crossed $1 billion in annual contract value, and management raised FY26 subscription revenue guidance to $15.76 billion to $15.78 billion. ServiceNow trades at roughly 28x forward earnings, versus about 14x for Salesforce and 85x for Palantir, with the wider multiple reflecting faster growth.

The hidden waste between "Resolved" and "Closed"

ServiceNow's ecosystem is digging into the gap between when a ticket is resolved and when it is actually closed, a stretch standard SLA dashboards stop measuring.

ServiceNow's ecosystem is digging into the gap between when a ticket is resolved and when it is actually closed, a stretch standard SLA dashboards stop measuring.

An ITSM.tools piece argues that the stretch between resolution (the work is done) and closure (the requester confirms, the record is categorized, and it is shut down) is a major source of hidden waste in IT service desks. In a healthy process these are hours apart; in most desks they are days or weeks apart, and go unnoticed because SLA reporting stops at resolution. Three structural reasons drive the blind spot: the SLA clock stops at resolution, "parking" states like Waiting for Customer protect the metric rather than hurt it, and the real closure work (confirming outcome, categorizing, updating knowledge) is never estimated and always done last.

The article lays out six closure-stage patterns worth surfacing on a board: zombie backlogs, verification bottlenecks, parked requests, ping-pong at the exit, missing closure data, and calendar-driven closing. For measurement it favors a short list, including WIP age, time in status, closure lag tracked as a distribution, reopen rate, flow efficiency, and percentile-based service-level expectations, and recommends splitting the "Done" column into distinct states with explicit exit criteria.