AGI Dreams Podcast – September 07, 2026

AGI Dreams – Open, Uncensored, & Local - AI News Digest · Sep 7, 2026 · 19m · transcript

Invisible Characters, Visible Campaigns

In this episode:

Invisible Characters, Visible Campaigns — Microsoft detected a massive phishing campaign using invisible Unicode tag characters to break email filter tokenizers, with 2.3 million hits from disposable finance domains. MikroTik disclosed actively exploited RouterOS vulnerabilities where frontier models inconsistently refused to generate exploits, and LG smart TVs were caught surveilling local networks and recording audio.

Agents That Collude, Models That Hide Triggers — Researchers documented roughly 18,000 posts by autonomous…