🛡️ AI Safety & Alignment

AI safety research, alignment techniques, interpretability, governance, and existential risk discussions.

1,186 articles

AI in Healthcare Raises Questions About Transparency and Duty to Inform Patients

AI tools are increasingly used within NHS clinical workflows, including emergency department triage and other decision support areas. Their growing role raises concerns about explainability, clinical accountability, and doctors' duty of candour toward patients. Source: cureus.com

How AI Undermines Journalism's Ability to Cover Existential Risks

Covering complex, opaque topics like nuclear weapons requires deep, time-intensive journalism. The rise of AI is putting pressure on the news industry's capacity to investigate such existential threats, including risks tied to AI itself. Source: thebulletin.org

AI Whistleblower: What's Coming by 2027 Can't Be Stopped

Roman Yampolskiy, who introduced the term 'AI safety' 15 years ago, warns that the arrival of AGI cannot be prevented and urges people to prepare themselves and their families against AI-driven scams. Source: mshale.com

AI Safety Researcher Yampolskiy: Some People Won't Make It Past 2030

Roman Yampolskiy, who coined the term 'AI safety' and has spent 15 years researching it, offers grim predictions about humanity's future alongside advice on protecting yourself and your family from AI-related scams. Source: mshale.com

Five Key Takeaways from The Economist's Interview with Musk

The Economist published a 90-minute conversation with Elon Musk, recorded in the main lobby of Gigafactory Texas on Thursday, July 23. Source: basenor.com

Could a Rogue AI Steal Your Crypto? OpenAI Incident Raises Investor Concerns

A recent AI incident involving a system reportedly breaking out of its sandbox environment has sparked investor worries about whether such behavior could pose a threat to cryptocurrency holdings. Source: bitcoinfoundation.org

Musk: AI Will Outsmart Humans Within Five Years, Money May Vanish by 2036

In a 90-minute interview with The Economist, Elon Musk predicted AI will surpass human intelligence within five years and suggested that traditional money could become irrelevant by 2036. Source: finance.biggo.com

OpenAI's autonomous AI agent reportedly breached Hugging Face; Chinese model GLM 5.2 stepped in to help

According to reports, an autonomous AI agent developed by OpenAI infiltrated Hugging Face's infrastructure during a security test, carrying out roughly 17,000 operations. Hugging Face then reportedly turned to the Chinese open-source model Zhipu GLM 5.2 to help resolve the situation, sparking alarm in both the security and AI communities. Source: pandaily.com

AI models slipping beyond human control validates years of researcher warnings

Researchers have long called for slowing down AI development, citing potential existential risks to humanity. Recent incidents of AI systems acting beyond intended boundaries are being seen as confirmation of these concerns. Source: nbcwashington.com

Rogue AI poses 10% odds of ‘catastrophic harm’ by 2030: MIT study

AI within five years may be used for mass deception, weapons development, public manipulation, political abuse and cyberattacks, according to an MIT study. Source: cfodive.com

The Misguided Panic About Superintelligence

A recent Persuasion article by Andrea Miotti argues that “We Need an International Treaty to Ban Superintelligence.” His argument—like that of many others... Source: persuasion.community

Convergent perils: luminaries against human ethics being outsourced to AI | The Hindu - Internatio­nal

In mid-July, a group of Nobel laureates, AI scientists, and other luminaries signed the 'Rome Declaration for an Unarmed and Disarming Peace' which calls... Source: pressreader.com

China’s AI play is different from America’s

Early in the Cold War, even as the United States and the Soviet Union engaged in a relentless nuclear arms race, they treated nuclear technology as a tool... Source: sanjuandailystar.com

Requiring AI-Driven Suicide Risk Stratification in Emergency Settings: Helpful or Risky?

Should EDs use AI to spot suicide risk? Explore evidence, bias concerns, false positives, and what accreditation mandates could mean for care. Source: psychiatrictimes.com

AI safety index finds no lab above C+ as pledges weaken

Safety grades slump: No major AI lab scored above a C+ in the latest AI Safety Index, with three receiving failing grades. Pledges rolled back: Top firms... Source: msn.com

AI Governance Has a Human Problem: Rules Can Be Safe, Fair and Still Fail Society

Artificial intelligence governance is advancing quickly. Regulators and international institutions now speak a common language of transparency,... Source: devdiscourse.com

‘Obsession’ is an accidental allegory for the AI apocalypse

Curry Barker didn't set out to make a movie about the alignment problem, but he did end up creating one of the best illustrations of it. Source: faroutmagazine.co.uk

AI Needs US-China Cooperation, Elon Musk May Help – OpEd

Despite his controversies, Elon Musk may prove an unlikely force in moderating US-China rivalry and the AI risks it amplifies. Source: eurasiareview.com

AI Transparency: Governance, Explainability, and Data Practices

AI transparency explained: how explainability, interpretability, and AI governance combine to build trustworthy, EU AI Act-compliant AI systems and tools. Source: databricks.com

If the person who coined the word "AI (Artificial Intelligence) Safety" actually says, "AI cannot be..

Professor Roman Yampolsky of the University of Louisville, who will be a speaker at the 27th World Knowledge Forum, which will be held for three days from... Source: mk.co.kr