Security and AI news from the week beginning 22 June 2026

Security and AI news from the week beginning 22 June 2026 highlighted major personnel shuffles and escalating cyber risks. Alphabet’s AI leadership was hit as Noam Shazeer left for OpenAI, while John Jumper and two other Gemini team members joined Anthropic. Over three days, the market reaction coincided with a £212 billion drop in Alphabet’s market capitalization. Gemini 3.5 Pro also missed its June general availability timeline for the second straight time, despite a 2-million-token context window and “Deep Think” reasoning. On the security front, OpenAI’s “Daybreak” deployed GPT-5.5-Cyber to audit open-source code and flagged 24 Linux kernel local privilege escalation exploits plus “Squidbleed,” a critical Squid proxy flaw hidden for 29 years. Meanwhile, Tenet Security reported “Agentjacking,” with 85% of tested AI coding agents vulnerable to hijacking via compromised Sentry data.