Breaking news and updates from the world of technology.
The Defense Department has formally authorized Amazon, Google, Microsoft, OpenAI, SpaceX, NVIDIA, Reflection AI, and Oracle to deploy their artificial intelligence systems on America's most sensitive classified networks—while a new White House directive mandates all defense agencies adopt multi-vendor AI within 120 days.
OpenAI has expanded Lockdown Mode — a security setting that disables web access, Agent Mode, Deep Research, and external connections — from enterprise customers to all ChatGPT accounts. It's the company's most direct response yet to the growing threat of prompt injection.
OpenAI's ChatGPT crossed 1 billion global monthly active users in May 2026, outpacing TikTok, Instagram, and YouTube to claim the fastest consumer app growth ever recorded. The milestone lands as the AI assistant race intensifies.
Security startup depthfirst ran an autonomous AI agent against FFmpeg's 1.5 million lines of C code and came back with 21 previously unknown vulnerabilities. The compute bill was $1,000. The oldest bug had been sitting untouched for 23 years. All 21 come with published proof-of-concept inputs.
A denial-of-service vulnerability in SolarWinds Serv-U lets unauthenticated attackers crash file transfer servers with a single malformed POST request. CISA added CVE-2026-28318 to its Known Exploited Vulnerabilities catalog on June 5 with a federal remediation deadline of June 19. The fix is in Serv-U 15.5.4 HF1.
A regulatory filing published Friday reveals Google has signed a $920 million per month cloud services agreement with SpaceX, giving it access to 110,000 Nvidia GPUs through June 2029. The deal positions SpaceX as one of the largest AI infrastructure providers in the world — and comes days before its anticipated IPO.
Security firm Calif has disclosed CVE-2026-49975, a denial-of-service vulnerability in HTTP/2 implementations that crashes major web servers using a single 100Mbps connection. Public exploit code appeared on GitHub on June 2.
CVE-2025-48595, a zero-click elevation-of-privilege vulnerability in the Android Framework, is being actively exploited in targeted attacks. CISA has added it to its Known Exploited Vulnerabilities catalog. Android 14, 15, and 16 are all affected.
CVE-2026-20245 allows attackers with netadmin access to escalate to root and push configuration changes to edge devices. Cisco has confirmed active exploitation and has no fix available yet.
The company behind Claude published a manifesto urging coordinated international restraint on the most powerful AI systems — citing the risk that models may soon be able to improve themselves faster than humans can oversee them.
A security researcher using Anthropic's Claude Opus 4.8 uncovered a critical soundness bug in Zcash's Orchard shielded pool that had gone undetected since May 2022, allowing the same shielded ZEC to be spent more than once. An emergency hard fork patched the circuit before any exploit was confirmed.