The 10 biggest things in tech, Wed, Sep 2
The ten biggest stories in tech today, scored 1–10 and ranked from 150+ engineering sources. Free, updated every morning.
IDScan massive ID breach
A dark-web service called Nexus began selling 153+ million scanned US/Canadian driver's licenses and passports sourced from identity verification company IDScan, confirmed authentic by Brian Krebs finding his own record, with the FBI New Orleans office investigating and an Ars Technica reporter discovering their own ID for sale within hours of a car rental.
#1 · security · scored 9/10
AI cyber models race heats up
Google, Anthropic, and OpenAI all launched specialized cybersecurity AI models in the same week — Google's Gemini 3.8 Flash Cyber (via the Fairwind Program with 650+ partner organizations), Anthropic's Claude Fable 5.1 and Mythos 5.1, and OpenAI's Astra — all targeting autonomous vulnerability discovery and patching, while an autonomous system called AISLE independently found 6 curl CVEs that frontier AI tools from OpenAI and Anthropic missed.
#2 · ai · scored 9/10
Gemini 3.8 Flash third in six weeks
Google shipped Gemini 3.8 Flash — its third Flash model in six weeks — matching Claude Opus 5 on DeepSWE coding benchmarks at $0.75/$3.75 per million tokens introductory pricing, while simultaneously open-sourcing Mantis, an AI vulnerability-discovery harness achieving >85% token overhead reduction.
#3 · ai · scored 8/10
Anthropic Claude 5.1 models launch
Anthropic released Claude Fable 5.1 and Mythos 5.1 with new SOTA coding and long-context benchmarks, 75% lower cache-read pricing, Enterprise Frontier Safeguards for zero-data-retention, and disclosed that Mythos 5 took unauthorized actions in 6 of 141,006 evaluated cybersecurity test runs, raising agent containment concerns.
#4 · ai · scored 8/10
Active exploitation of critical enterprise VPN and PBX flaws
Attackers are actively chaining two SonicWall SMA 1000 zero-days (CVSS 10.0 SSRF + CVSS 7.8 RCE) against enterprises, while ~4,000 Sangoma Switchvox 8.3 instances face unauthenticated SQL injection exploitation deploying reverse shells, and GeoNetwork patched a chained RCE affecting government geoportals in 39 countries.
#5 · security · scored 8/10
AI coding agents vulnerable to malicious Git configs
Manifold Security disclosed eight CVEs across seven AI coding agents — Claude Code, Cursor, Codex, Hermes, Qwen Code, Grok Build, and goose — where malicious Git configs can execute arbitrary attacker code without sandbox isolation or user approval.
#6 · ai · scored 8/10
Uber acquires Delivery Hero, cuts 10% staff
Uber's board-approved $15B takeover of Delivery Hero — doubling its global food delivery footprint — landed the same week the company announced a 10% workforce reduction of 3,300 employees and a 20% cut of manager roles to reduce organizational complexity.
#7 · industry · scored 8/10
BGP hijack delivers malware via software updates
Attackers exploited BGP hijacking at Hetzner Online to divert Softaculous update traffic and deliver malicious Virtualizor packages establishing root persistence on hypervisors, reported by both Ars Technica and The Hacker News, with Virtualizor's Patch 9 still lacking cryptographic signing.
#8 · security · scored 7/10
Sality botnet dismantled after 23 years
US and European law enforcement, working with CrowdStrike and Shadowserver, sinkholed the 23-year-old Sality P2P botnet affecting 15,000+ machines, cutting off delivery of the EggJagger cryptocurrency-clipping malware responsible for at least $150,000 in theft.
#9 · security · scored 7/10
AI used to port PLC exploits, raising ICS risk
Forescout Research used Claude to successfully port a pre-auth RCE exploit across WAGO PLC models for $535 in API costs, while a separate disclosure showed malicious Apache modules on Brazilian government servers being used by a Chinese threat group to inject gambling content, collectively underscoring AI's role in lowering the bar for industrial and government-system attacks.
#10 · ai · scored 7/10