5 PM Technology News Recap: FBI Jobs Site, Alibaba Cloud, Surface, Comma, and Youth AI Safety (September 23, 2026)

The FBI investigates unauthorized activity affecting its recruiting site; Alibaba plans new cloud regions; Microsoft refreshes smaller Surface PCs; NHTSA opens a preliminary comma.ai evaluation; and an ex-Google safety leader calls for independent youth-AI standards.

Reader ratingRate this article

Loading reader rating…

Privacy-conscious voting uses a signed random browser identifier and a short-lived network hash; the rating tables contain no raw IP-address column.

Wednesday’s 5 PM technology recap separates five real developments from their loudest claims: the FBI is investigating unauthorized activity affecting its public recruiting site; Alibaba Cloud announced a multinational expansion plan; Microsoft refreshed its smaller Surface PCs; federal safety investigators opened a preliminary evaluation of comma.ai devices; and a former Google trust-and-safety leader called for independent youth-AI standards.[1][3][5][8][10]

Reporting cutoff: September 23, 2026, 5:10 p.m. EDT (21:10 UTC). The FBI has confirmed an investigation, not the attackers’ claimed data scope. Alibaba’s new regions are plans, not operating capacity. Microsoft’s performance percentages are company benchmarks. NHTSA’s comma.ai matter is a preliminary evaluation, not a defect finding. The comparative warning about AI and children is an expert’s assessment, not a new population-level causal study.

Listen to the broadcast-ready recap

This KittenTTS Bella narration was synthesized locally from the same checked facts as the article.

1. FBI investigates activity affecting its recruiting site

The FBI said it is investigating “unauthorized activity affecting FBIjobs.gov” after ShinyHunters defaced the recruiting site and claimed to have stolen information about employees and applicants.[1] As of Wednesday morning, The Record found the Special Agent Applicant Portal unavailable.

The group gave journalists a purported sample of 5,000 records. The Record said multiple outlets confirmed the legitimacy of samples, but the FBI has not confirmed the attackers’ claimed scope or that the material came from internal FBI systems.[1][2]

How we know—and what remains unknown: the agency acknowledges an investigation into the recruiting site. The threat actor’s claim that it holds information on every employee and applicant remains unverified, and a sample containing real-looking records does not prove the claimed source, completeness, or current accuracy.

Practical actions: current or former applicants should be skeptical of unexpected messages that cite job applications, personnel details, or urgent account verification. Use independently located FBI contact channels rather than links or phone numbers in a message, and preserve suspicious communications for reporting.

Good: public acknowledgment and outside sample checking narrow the uncertainty faster than silence would.
Caution: exposed recruiting or personnel data could enable convincing impersonation even if the broader breach claim is overstated.

2. Alibaba Cloud plans three new regions and wider capacity

Alibaba Cloud announced that it plans to establish its first cloud regions in Türkiye, Finland, and the Netherlands within 12 months, while expanding capacity in Malaysia, Germany, the United Arab Emirates, France, and Hong Kong.[3][4] The company says it currently operates 107 availability zones across 31 regions.

Alibaba also introduced Smart Studio for packaging and billing model APIs, Smart Fusion for multi-model routing, and Smart Video for prompt-driven video creation.[3] Its claims of up to 505% more inference throughput and roughly 50% lower token spending are company-reported and do not include enough public methodology here to treat them as general guarantees.

Plan versus installed capacity: the region list and timeline are announced intentions. A planned region should not be treated as production-ready until Alibaba publishes launch status, service availability, data-location terms, and resilience details for that region.

Practical actions: enterprises considering a new region should verify service parity, data-residency contracts, cross-region backup options, exit costs, support coverage, and measured application performance before migrating regulated or latency-sensitive workloads.

Good: more regional choices may reduce latency and give organizations additional data-location options.
Caution: geography alone does not establish legal compliance, availability, or operational independence.

3. Microsoft refreshes its smaller Surface PCs

Microsoft announced second-generation Surface Pro 12-inch and Surface Laptop 13-inch models using six-core Snapdragon X2 Plus processors. Both start with 16 GB of memory, use brighter 500-nit displays, and are scheduled for October 13 availability in selected markets.[5][6]

The Surface Pro starts at $1,149.99 and the Surface Laptop at $1,199.99; Microsoft also announced a $79.99 haptic Surface Mouse.[5] The Verge noted that removing the former 8 GB base configurations explains part of the higher entry prices.[6]

Microsoft advertises up to 18% more efficient battery performance, more than 60% faster graphics, and up to 95% faster local AI inferencing compared with prior models. Those results come from Microsoft’s preproduction testing and named benchmarks; real gains will vary by workload, settings, and configuration.[5]

Buyer context: this is primarily a processor, memory-floor, brightness, and battery refresh—not a complete redesign. The devices remain Arm-based Windows PCs, so buyers with specialized drivers, older peripherals, virtual machines, or niche business software should test compatibility.

Practical actions: compare the price of a complete Surface Pro setup, including its separately sold keyboard or pen if needed. Business buyers should validate their security tools, VPN, printers, line-of-business software, and repair/support plan before standardizing.

4. NHTSA opens a preliminary comma.ai safety evaluation

NHTSA’s Office of Defects Investigation opened Preliminary Evaluation PE26007 on September 21 after identifying five crashes in which vehicles with comma.ai devices reportedly struck stopped or slow-moving vehicles in the same lane.[8] Two crashes resulted in three deaths; four crashes contributed to nonfatal injuries to 11 other people.

The agency says preliminary information suggests the systems may not have adequately detected or responded to hazards, but it has only confirmed that the system was powered on or engaged in multiple crashes—not necessarily all five.[8] Investigators will examine system status, performance, safeguards, and openpilot-derived software forks. Ars Technica brought the investigation into today’s news window.[7]

Legal and technical state: a preliminary evaluation is an investigation, not a conclusion that a defect exists or that the system caused every crash. Modified forks further complicate software-version and responsibility questions.

Practical actions: users should keep the driver engaged, follow the manufacturer’s instructions, avoid unsupported software forks, document the installed hardware and exact software version, and report safety incidents to NHTSA. No driver-assistance device makes a vehicle autonomous.

Good: the evaluation can assemble crash data and distinguish hardware, official software, and modified-fork behavior.
Caution: stopped-object detection is a high-consequence edge case; drivers cannot outsource attention to an aftermarket aid.

5. Former Google safety leader urges independent youth-AI standards

Tom Siegel, the founding leader of Google’s trust-and-safety team and the new executive director of Common Sense Media’s Youth AI Safety Institute, told Reuters that AI companies need stronger age assurance, more child-focused controls, and independent “crash test”-style evaluations.[9][10]

Siegel warned that AI could harm children more severely than social media did and cited risks including suicide, psychosis, and reduced critical thinking. That is his risk assessment—not a demonstrated comparative population-level finding.[10] Google and OpenAI described existing parental controls and age-detection measures; Anthropic did not immediately respond to Reuters.

What this story is: an experienced safety leader’s policy and product argument, supported by company responses. It is not a new clinical study, a regulator’s finding, or proof that every AI service creates the same risk.

Practical actions: families and schools should define which tools are allowed, review privacy and memory settings, keep high-stakes emotional or medical conversations with trusted adults and qualified professionals, and teach children that fluent answers can still be wrong or manipulative.

Good: independent test standards could make safety claims easier to compare than self-selected company demonstrations.
Caution: age assurance can introduce privacy tradeoffs, and voluntary standards work only when methods and results are transparent.

What to do next

  1. Applicants and employees: verify unexpected FBI-related outreach through independently found channels.
  2. Cloud teams: treat new-region announcements as planning signals until service, contract, and resilience details are live.
  3. PC buyers: test Arm software and peripheral compatibility before adopting the new Surface systems.
  4. Drivers: keep full responsibility and avoid unsupported driver-assistance software forks.
  5. Families and schools: pair AI access with clear rules, privacy review, adult escalation, and independent safety evidence.

Need help securing, fixing, or planning your technology? Schedule The IT Guys

Frequently asked questions

Did hackers breach the FBI’s internal systems?

The FBI confirmed an investigation into unauthorized activity affecting FBIjobs.gov. It has not confirmed the group’s broader claims or that the sampled records came from internal FBI systems.[1]

Are Alibaba’s new cloud regions online now?

No. Alibaba said the first regions in Türkiye, Finland, and the Netherlands are planned within 12 months. Verify live service status and contracts before making a production decision.[3]

Do the new Surface models guarantee the advertised performance gains?

No. Microsoft’s percentages come from specific preproduction tests. Real results depend on configuration, workload, settings, and software compatibility.[5]

Has NHTSA found comma.ai devices defective?

No. PE26007 is a preliminary evaluation. NHTSA is gathering evidence about incidents, system status, performance, safeguards, and software forks.[8]

Did a new study prove AI harms children more than social media?

No. That comparison is Tom Siegel’s warning in a Reuters interview, not a new population-level causal study.[10]