Tech news · 15 November 2026

Tech news, 15 November 2026: the Fireworks AI breach spreads to Cursor and Uber, phishing follows the Microsoft outage, Check Point's exploited zero-day and SpaceX's AI campuses

The Fireworks AI breach is now showing up in its customers' disclosures: Cursor traced its own incident to a key it had given Fireworks, and Uber said driver records held by Fireworks were taken. Attackers used the Microsoft 365 outage as a phishing lure, a Check Point zero-day was exploited to gain root on gateways, and SpaceX announced plans for large AI computing campuses.

Tech newsSumanta Mukhopadhyay6 min read7 sources
#news#security#ai#platform
TECH NEWS · TECH NEWS · 15 NOVEMBER 202615 Nov 2026 · 08:50 GST

Today's brief covers stories from 12 to 14 November.

  1. VendorAn AI inference provider is breached
  2. KeyA customer's access key held by the vendor is exposed
  3. BucketThe key opens shared storage holding records with live credentials
  4. PivotThose credentials reach the customer's own services
  5. TamperThe attacker plants backdoored software for the customer's users
  6. RippleOther customers disclose their own data losses
How one vendor breach became several customer incidents, as described in Cursor's account.

1. The Fireworks AI breach reaches Cursor and Uber

Cursor said its October breach began with an access key it had issued to Fireworks AI being exposed; the key gave read-only access to a shared storage bucket whose records still contained credentials, which the attacker used to reach some of Cursor's services and plant backdoored extensions1. Separately, Uber said driver records held by Fireworks AI, including driver's licence information, were taken, according to state breach filings2.

CTO take: this is what supply chain risk looks like in AI: one provider's breach becomes many customers' incidents. Never give an AI provider a key with more reach than it needs, never let logs or stored records keep live credentials, and know which of your data each provider holds so you can respond the day they disclose. See AI inference providers are supply chain.

2. Microsoft restores 365 services and warns of outage-themed phishing

Microsoft said Microsoft 365 services were restored for all remaining affected customers after the multi-day disruption, and warned of a phishing campaign exploiting it3. The emails imitated service-health notices, cited a real incident number and pressured recipients to follow fake email recovery links4.

CTO take: every big outage is followed by phishing that uses it. Tell staff that real status information comes from the admin centre or your IT team, never from an email link, and check your mail filters for the reported patterns. See the Microsoft 365 outage playbook.

3. A Check Point zero-day is exploited to take over gateways

CISA added CVE-2026-123039, a privilege escalation flaw in Check Point's Gaia OS, to its Known Exploited Vulnerabilities catalogue on 13 November6. BleepingComputer reports attackers combined it with other flaws to gain root on gateways through the Mobile Access portal and deploy a wiper aimed at VMware ESXi5.

CTO take: security gateways keep appearing as entry points this year, from Fortinet to Check Point. Apply the vendor's hotfix, disable portals you do not use, check gateways for signs of compromise, and make sure backups of virtualisation hosts are isolated from the networks a compromised gateway can reach. See the management plane hardening case study.

4. SpaceX plans more than $74 billion of AI computing campuses

SpaceX announced plans for AI supercomputing campuses in Arizona and Texas worth more than $74 billion, with the Texas site to generate its own electricity, The Next Web reports7. The plans still depend on local approvals7.

CTO take: new AI capacity is increasingly built with its own power to avoid grid limits and local opposition. For buyers, more capacity is coming, but late and unevenly; plan with several providers and regions. See planning around the AI capacity crunch and the US page.

What I'd do this week

  • Review keys and data you have given AI providers, and remove anything not needed.
  • Warn staff about outage-themed phishing and where real status updates come from.
  • Patch Check Point gateways and disable unused portals.
  • Check that virtualisation backups are isolated from gateway networks.

I am based in India, the United States and the UAE, working globally. See where I work, or analyse your challenge and bring the result to a call.

References
  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
FAQ

Common questions

How was Cursor affected by the Fireworks AI breach?

Cursor said an access key it had issued to Fireworks AI was exposed, giving read-only access to a shared storage bucket containing records with credentials that had not been removed. The attacker used those to reach some of Cursor's software services and planted backdoored extensions.

What phishing followed the Microsoft 365 outage?

Microsoft warned of a credential-phishing campaign imitating service-health notices about the 12 November outage, including a real incident number, and pressuring recipients to follow fake email recovery links.

What is the Check Point flaw CISA added to its catalogue?

CVE-2026-123039 is a privilege escalation flaw in Check Point's Gaia OS, added to CISA's Known Exploited Vulnerabilities catalogue on 13 November 2026. Reporting says attackers combined it with other flaws to gain root on gateways.

Keep reading

Related

ServiceAI & LLM architectureI design and build AI systems that leadership can trust: grounded in your data, measured by evaluations, governed by your rules and priced so they make sense at scale.→ServiceFractional CTO and AI architect for US companiesSenior technology leadership for US founders, enterprises and investors: AI that reaches production, data platforms that pass security review, and governance that keeps pace with federal and state rules.→InsightTech news, 10 November 2026: Claude Fable 5.2 and Mythos 5.2, N-able takes N-central offline, Qubole's AWS access abused, Norway pauses data centre grid bookings and Meta's Muse EmberAnthropic released new Claude models with a lower price for its general model. A remote management platform used by managed service providers went offline after a researcher published alleged zero-days, and attackers used a data platform's trusted access into its customers' AWS accounts. Norway paused new grid reservations for data centres, and Meta launched a small, cheap model aimed at agent work.→InsightTech news, 9 November 2026: secrets in Claude Code settings files, AI buyers squeeze chip supply, Fortinet advisories due and Anthropic's IPO timingA quiet weekend, with a few stories that matter for engineering teams. Researchers found real credentials committed to public GitHub repositories through an AI coding tool's local settings file. A chipmaker said AI customers were taking capacity everyone else needs. Fortinet's advisories for already-released critical fixes land on Tuesday, and investors are questioning the timing and price of Anthropic's listing.→
Open to advisory, fractional CTO & select builds

Let's talk about what you're building.

The fastest route is a free 30-minute call. Prefer to write? Send a message and I'll reply personally within two business days.

Start a conversation

How would you like to reach me?

Book a 30-min callFastest. Pick any open slot on my calendar.→ Write in GmailOpens a pre-filled draft in your browser.→ Use my email appOutlook, Apple Mail or your default client.→

I reply personally within two business days. Time zones from the Americas and Europe to the Middle East and Asia-Pacific.