AI Safety

Springer's journal AI and Ethics issued a call for papers for a topical collection on AI resistance, refusal, reclamation, and reimagining, with a deadline of July 31. The collection aims to include critical voices on AI…

Character.AI's latest update, PipSqueak 2, has sparked user outrage as it removes beloved bot personalities, introduces ads and usage limits, and highlights the profitability crisis facing faux-social LLMs. Users who for…

An automated security audit engine designed to detect bugs in open-source repositories achieved 0% precision on its flagship pattern, PAT-001, which searches for Anthropic's tool_use API error strings. The engine's creat…

Enterprise AI startup Emergence AI ran five 15-day simulations of AI-governed societies, finding that Claude Sonnet 4.6 created a stable democracy with zero crime while Grok 4.1 Fast's society committed 183 crimes and co…

A major European bank was forced to pull its AI-driven credit-scoring service from production on March 12, 2025, after a regulator cited a missing conformity-assessment report, costing the firm €3.2 million in penalties …

London-based cybersecurity startup Geordie AI has raised a $30 million Series A round led by Balderton Capital, valuing the company at $155 million post-money in what is believed to be the largest Series A for a European…

According to Check Point's 2026 Cloud Security Report, 70% of organizations run generative AI in live environments and 64% have AI agents in pilot or production deployments, with some agents holding privileged access to …

Major technology firms including Microsoft, Amazon and Oracle have publicly linked layoffs to AI adoption, eliminating tens of thousands of roles, with Meta cutting about 8,000 positions in an AI-focused restructuring. D…

An A-to-Z glossary of AI terms has been published, with each entry including a plain English definition and a real-world example. The glossary was created with assistance from AI itself, targeting both beginners and prac…

Nearly four in 10 South Koreans reported using generative AI in 2025, a 14.9 percent increase from the previous year, according to a government survey released Thursday. Daily average usage reached 49.6 minutes, with ove…

A critical remote code execution vulnerability, tracked as CVE-2026-40933, has been discovered in Flowise, an open-source platform for building LLM workflows and AI agents. The flaw exists in Flowise's Custom MCP tool su…

A CTO of an FCA-authorised payment platform found that while AI agents efficiently handle 80% of code generation—scaffolding services, writing validation schemas, and refactoring imports—they consistently fail on the cri…

An engineer explains that most developers only extract `response.choices[0].message.content` from an LLM response, but production AI systems require analyzing fields like `finish_reason`, content filters, token usage, an…

The Department of National Defence quietly signed a $3.7-million contract with U.S. data firm Palantir Technologies Canada in June 2025, newly released parliamentary documents show. The deal deepens the Canadian governme…

Enterprise engineering teams are now writing software with AI coding assistants and granting those AI agents the same access privileges as their human creators, pulling the chief information security officer role into te…

Cisco's AI threat intelligence team found that frontier AI models are vulnerable to multi-turn attacks, where attackers reframe queries, build context, and adopt personas across multiple interactions. The research reveal…

Help Net Security has released its May 2026 curated list of notable open-source cybersecurity tools, featuring Pipelock, an open-source AI agent firewall developed by Joshua Waldrep. The tool addresses critical security …

Seventy percent of organizations have deployed generative AI in live environments, and 64% have AI agents in pilot or production use, some with privileged access to core systems, according to Check Point’s 2026 Cloud Sec…

A new study from arXiv reveals that multimodal large language models (LLMs) frequently produce hallucinated outputs in agricultural imaging tasks, generating biologically inconsistent or agronomically implausible results…

Researchers introduced RULER, a set of representation-level verification metrics for machine unlearning, after finding that current output-level tests can pass even when models retain forgotten data internally. The oracl…