LLM-Generated Passwords Look Strong but Crack in Hours, Researchers Find

msmash / slashdot - AI security firm Irregular has found that passwords generated by major large language models -- Claude, ChatGPT and Gemini -- appear complex but follow predictable patterns that make them crackable in hours, even on decades-old hardware. When researchers …

#ai #cybersecurity #ml #dataprivacy #infosec #genai #chatgpt #cybercrime #hack #encryption

AI vibe-generates the same ‘random’ passwords over and over

15 days / pivot-to-ai / David Gerard

15 days / slashdot / msmash

AI-generated passwords are surprisingly easy to crack, researchers find

15 days / techspot

AI-Generated Passwords Are Apparently Quite Easy to Crack

15 days / gizmodo / AJ Dellinger

New benchmark shows AI agents can exploit most smart contract vulnerabilities on their own

Maximilian Schreiner / the-decoder - OpenAI and crypto investment firm Paradigm have built EVMbench, a benchmark that measures how well AI agents can find, fix, and exploit security vulnerabilities in Ethereum smart contracts.The article New benchmark shows AI agents can exploit most smart c…

#blockchain #cybersecurity #crypto #openai #aiethics #genai #cybercrime #developertools #wallet #research

OpenAI Just Showed That AI Can Drain a Crypto Wallet… on Purpose

15 days / eweek / Grant Harvey

New benchmark shows AI agents can exploit most smart contract vulnerabilities on their own

15 days / the-decoder / Maximilian Schreiner

Morning Minute: OpenAI and Paradigm Turn Focus to Smart Contracts

15 days / decrypt / Tyler Warner

CYBERCRIME Heatmap

90 days, weeks are vertical, left is older; hover for info, click to see that day's coverage.

More Top Stories...

Pentagon labels Anthropic a supply‑chain risk; company vows legal fight

The Pentagon has designated Anthropic and its products as a “supply‑chain risk,” prompting the company to announce a court challenge. Experts warn the move could chill collaboration and talent flows into AI, while Anthropic insists it will contest the determination to protect its operations and customers. More...

Oracle and OpenAI scrap Texas data-center expansion; Meta eyes the spare capacity

Oracle and OpenAI have abandoned plans to expand a flagship Texas data center, leaving substantial compute capacity up for grabs. Nvidia reportedly brokered interest from Meta to take the unused slots as OpenAI downscales that particular buildout, a move that rattled markets and highlights shifting demand for large-scale on-prem AI infrastructure. More...

SoftBank seeks massive $40B loan to back OpenAI investment, courting big risk

SoftBank is reportedly seeking up to a $40 billion loan to finance its planned stake in OpenAI, an audacious use of leverage to double down on the AI boom. The move would be one of the largest single‑company financing gambits in recent memory, raising questions about balance‑sheet strain versus potential upside. More...

OpenAI launches Codex Security agent to automatically detect software vulnerabilities

OpenAI rolled out Codex Security, an AI agent that scans codebases to find complex vulnerabilities, suggests actionable fixes, and uses sandbox testing to limit false positives. The tool has already flagged issues in major projects and aims to compete with traditional application security tooling by automating deep, contextual code review. More...

Google releases Workspace CLI enabling AI agents to access Gmail, Drive, Calendar

Google has shipped an open-source Workspace CLI that gives AI agents like OpenClaw programmatic access to Gmail, Drive, Calendar and other Workspace services via a built-in MCP server. The tool standardizes agent integration, making it easier — and slightly creepier — for automated assistants to act on users’ behalf across core productivity apps. More...