☕️ FBI reports record $21 billion lost to cybercrime last year

Q: Anthropic developed Claude Mythos Preview AI that can autonomously exploit vulnerabilities but won't

Anthropic developed Claude Mythos Preview AI that can autonomously exploit vulnerabilities but won't release it publicly due to security risks.

Q: FBI reports Americans lost record $21 billion to cybercrime in 2025, with AI-related scams causing $

FBI reports Americans lost record $21 billion to cybercrime in 2025, with AI-related scams causing $893 million in losses.

Q: Google AI Overviews delivers wrong answers 10% of the time, potentially producing millions of incorr

Google AI Overviews delivers wrong answers 10% of the time, potentially producing millions of incorrect responses daily.

Techpresso·Wednesday, April 8, 2026·1 min read

Technology AI/ML Security

Share𝕏 in

AI Summary

This technology newsletter covers major AI and cybersecurity developments, including Anthropic's dangerous new AI model that can exploit vulnerabilities, record FBI cybercrime losses of $21 billion, and Google's AI accuracy issues. The newsletter also covers product updates like Chrome's vertical tabs and claims about Bitcoin's creator identity.

Key Facts

✓Anthropic developed Claude Mythos Preview AI that can autonomously exploit vulnerabilities but won't release it publicly due to security risks.

✓FBI reports Americans lost record $21 billion to cybercrime in 2025, with AI-related scams causing $893 million in losses.

✓Google AI Overviews delivers wrong answers 10% of the time, potentially producing millions of incorrect responses daily.

Author Takes

BearishTechpresso

Google AI accuracy

Even 91% accuracy means hundreds of thousands of lies going out every minute

More from Techpresso

☕️ Alexa is moving into Amazon.com

Google unveiled Googlebooks, a new line of AI-native laptops powered by Gemini and built on Android, succeeding Chromebooks after 15 years. Amazon is

May 13

☕️ eBay rejects GameStop's takeover offer

This edition of Techpresso covers eBay rejecting GameStop's $56 billion unsolicited takeover bid, OpenAI launching a new cybersecurity project called

May 12

☕️ OpenAI launches an AI consulting arm

This edition of Techpresso covers OpenAI launching a $4B AI consulting arm called DeployCo, Google detecting the first AI-built zero-day exploit, and

May 11