Google denies ‘misleading’ reports of Gmail using your emails to train AI

Google pushed back on viral social posts claiming Gmail content is being used to train its Gemini AI, saying Smart Features predate the reports and that it does not use Gmail messages or attachments to train its AI models; the company noted a January settings update that separated Workspace and other product personalization, though a Verge staffer reported being re‑opted into some features, highlighting possible UX/consent friction. Separately, a sponsored piece citing Cisco research reported that 91% of organizations are increasing network investments for AI workloads, 71% say current data‑center capacity is insufficient and 88% are adding capacity, and Cisco is pitching management and AgenticOps tools to address these bottlenecks. Taken together, the items narrow an immediate privacy/regulatory overhang for Alphabet while underscoring a potential capex tailwind for networking and data‑center vendors as enterprises upgrade infrastructure for generative AI.

Analysis

Google publicly denied viral claims that it is using the content of Gmail messages and attachments to train its Gemini AI model, with spokesperson Jenny Thomson stating “we do not use your Gmail content for training our Gemini AI model” and noting Gmail Smart Features predate the reports. A January settings change separated personalization controls for Google Workspace from other Google products, and at least one Verge staffer reported being re-opted into some Smart Features, highlighting potential UX or consent friction despite Google’s denial.

A direct corporate denial reduces an immediate privacy/regulatory overhang for Alphabet relative to the viral narrative, but the reported re‑opt‑in incident could sustain reputational scrutiny and motivate closer regulatory or user attention to consent flows. Separately, Cisco‑sponsored research shows 91% of organizations are increasing network investments for AI, 71% report insufficient data‑center capacity, and 88% are expanding capacity, with Cisco pitching AgenticOps and other networking tools as solutions.

Taken together, the item narrows short‑term privacy risk for Google while reinforcing a visible enterprise infrastructure capex tailwind for networking and data‑center vendors; market signals are mildly positive (sentiment score 0.22, market impact 0.35) but operational and consent execution remain watch points.

Key Decisions for Investors

• Reassess near‑term regulatory risk for Alphabet and consider maintaining current exposure rather than initiating material reductions, while monitoring any follow‑up admissions, policy changes, or mass re‑opt‑in reports
• Increase selective exposure to networking and data‑center infrastructure vendors that benefit from enterprise AI capex (Cisco and peers referenced in the study), but size positions to account for execution risk and procurement lead times
• Monitor corporate consent/UX metrics, privacy disclosures, and vendor order/backlog updates as leading indicators; adjust positions if evidence emerges of broader consent failures or regulatory actions that could reverse the sentiment tailwind