Digital Banking Safety

Block’s Dessa speech model laughs and interrupts in company demos, but has no public release, independent testing or published benchmarks.

• CheckingMarket Team
• October 26, 2026

Recent agent incidents show how AI tools can move tokens, expose private data or access accounts—and why permissions and contracts matter.

• CheckingMarket Team
• October 25, 2026

UK and Anthropic reports detail how Mythos 5 targeted real developers during a cyber test, why it continued, and what remains uncertain.

• CheckingMarket Team
• October 25, 2026

Anthropic launched Claude Duet as a waitlisted research preview, with approval rules and limits that remain untested independently.

• CheckingMarket Team
• October 25, 2026

US agencies and Microsoft report AI-assisted attacks, while OpenAI says AI is finding vulnerabilities defenders can patch.

• CheckingMarket Team
• October 24, 2026

Railway says an attacker may have accessed some customer environment variables. The company has not named affected keys or confirmed their use.

• CheckingMarket Team
• October 22, 2026

Two cases show how agents can pass messages and act on them, raising questions about authorization, security and oversight.

• CheckingMarket Team
• October 18, 2026

A large AI-generated mathematics release has renewed debate about cryptographic assumptions, but no cryptographic system has been shown to be broken.

• CheckingMarket Team
• October 9, 2026

A new AI management benchmark reveals even the weakest AI managers score 26 points, highlighting the importance of trust and task completion in AI performance.

• CheckingMarket Team
• September 19, 2026

A user reports that spending $220 on Google app ads resulted in 60% of installs being from bots, raising concerns about ad fraud and app marketing integrity.

• CheckingMarket Team
• September 12, 2026