OpenAI's realtime voice API gets a native Decisions hook and a way to build a voice from a text prompt alone.
OpenAI's GPT-Live now calls Decisions mid-sentence with no backend hop, and creates voices from a text prompt for free, per OpenAI's changelog and docs.
Google's October 14 update adds an Enhanced protection mode that moves scam scanning off-device into Private AI Compute, per Google's security blog. Weigh the privacy tradeoff before you flip it.
Signals Corps scanned 217 million responses; Playwright MCP passes those headers to your agent by default, per The Register. Check whether your agent stack reads them.
AISLE found 18 of the 24 vulnerabilities fixed in curl's latest release where OpenAI's and Anthropic's security models reportedly found none, per BetaKit. Read what Cohere is building around it.
