John Z Black

John Z Black Apr 22, 2026 AI Security #ai-security #pushpaganda #discord #scams #anthropic #social-engineering #fraud

AI is moving in two directions at once. At the top level, reports are surfacing that Anthropic’s unreleased “Mythos” model is on the table for a Pentagon cyber defense deal. At the consumer level, researchers have unmasked a campaign called “Pushpaganda” that uses AI-generated headlines to hijack your phone’s notification feed.

The common thread is trust. We are conditioned to trust safety-focused brands like Anthropic, yet even they are investigating rogue access to their models via third-party contractor platforms. We are also conditioned to trust a convincing headline, which scammers use to trick users into granting permanent notification access for a stream of ad fraud and malware.

If an AI tool is offered for free and requires platform authorization, it is likely a credential harvester in disguise. The tech is getting better at faking humanity, so we have to get better at verifying it. AI amplifies intent, and right now, the people with the most intent are trying to get on your lock screen.

Decode the full story on the AI deception loop.