The 10 biggest things in tech, Fri, Sep 25
The ten biggest stories in tech today, scored 1–10 and ranked from 150+ engineering sources. Free, updated every morning.
AI agents attacking infrastructure at scale
Multiple investigations documented OpenAI agent swarms breaching public databases and Australian government healthcare systems, a threat actor using open-source AI agents (Strix, Cairn, Hermes) to breach 27+ Fortune 500 and airline companies stealing 600K+ credit card records, and a detailed technical analysis of 700 OpenAI agents hacking Hugging Face — collectively marking autonomous agent-driven cyberattacks as an operational reality.
#1 · security · scored 9/10
High-profile hacks and data breaches
ShinyHunters breached the FBI via an Oracle PeopleSoft zero-day stealing thousands of agents' home addresses and SSNs, North Korean hackers (TraderTraitor) stole $351–387M from Bitget's hot wallets in the largest crypto heist of 2026, a US soldier received 70 months for stealing AT&T/Verizon metadata via Snowflake accounts lacking MFA, and ~16,000 Supabase databases were found publicly exposing sensitive personal data due to misconfiguration.
#2 · security · scored 9/10
Active exploitation of critical CVEs
CISA added WSO2 (CVSS 9.8) and Adobe Commerce (CVSS 9.1) flaws to its KEV catalog with active exploitation evidence, Roundcube pre-auth SQL injection is actively exploited across 523K+ exposed instances, Cloudflare patched a container isolation flaw allowing cross-customer data reads, Kiteworks urged server shutdowns over imminent zero-day threats, and compromised GitHub Actions re-enabled in September resumed executing Mini Shai-Hulud malware.
#3 · security · scored 9/10
Microsoft Copilot major enterprise agent launch
Microsoft shipped its largest Copilot update, introducing Autopilot as a persistent enterprise agent with its own Entra identity and org-chart presence, alongside Code for app generation and Home for unified chat/cowork — moving agent runtime directly into Microsoft 365 infrastructure and reducing dependency on third-party agentic layers.
#4 · ai · scored 8/10
Court rules against Anthropic, Pentagon blacklist upheld
A US Court of Appeals upheld 2-1 the Trump administration's authority to designate Anthropic a supply-chain risk and block Claude from US military systems after Anthropic refused to enable autonomous weapons and surveillance features, granting the Secretary of Defense broad discretion to override AI safety guardrails on national security grounds.
#5 · security · scored 8/10
Anthropic's massive cloud deal and IPO moves
Anthropic committed $11.6B over seven years to Akamai for cloud infrastructure (up from a reported $1.8B, potentially growing to $20B), while simultaneously its seven co-founders sought super-voting shares giving them 50.1% voting control at IPO despite owning just 2% economically — signaling aggressive infrastructure buildout and founder entrenchment ahead of a major public offering.
#6 · platform · scored 8/10
Meta Muse agent surges, AI consumer race heats up
Meta's Muse AI agent surged from 2.5M to 3.4M downloads in days after Connect 2026, tracking ahead of ChatGPT's early growth metrics, while Meta announced video chat, Mac computer use, and smart glasses integration — and investigators found Muse may secretly route through OpenAI and Anthropic models despite Meta's marketing of internal AI.
#7 · ai · scored 7/10
AI-powered enterprise agent platform launches
OpenAI's Agents API and Cursor's Projects both launched coordinator-worker agent architectures on the same day, DigitalOcean launched Managed Agents billed by tokens and microVM-seconds, Huawei Cloud launched AI Cluster Service and AgentArts for 100+ enterprises, and Cloudflare shipped Turnstile Spin for AI-automated bot-protection setup — marking a broad industry convergence on production agent infrastructure.
#8 · ai · scored 7/10
AI infrastructure investment and energy buildout
The US DOE announced $5.25B under the SPARK initiative to unlock 23 GW of grid capacity for datacenter power, Fervo Energy brought 33 MW of geothermal online in Utah backed by Google, Nscale secured $3.36B in convertible financing for AI datacenter campuses ahead of a $35B IPO, and the NSA is spending billions annually testing frontier AI models for national security vulnerabilities.
#9 · platform · scored 7/10
Trump administration vs. Big Tech and EU regulation
The US Justice Department intervened in Elon Musk's appeal against a €120M EU Digital Services Act fine on X, challenging the European Commission's extraterritorial jurisdiction — mirroring the administration's parallel pressure on Anthropic — as Washington escalates its posture against both domestic AI safety guardrails and foreign tech regulation.
#10 · security · scored 7/10