AI Safety
Follow AI Safety in the Factlen app and get an alert when there is news. Get the app
- All Technology
- Gadgets & Hardware
- Software & Platforms
- Apps & Mobile
- Cybersecurity & Privacy
- Cloud Computing
- Consumer Technology
- Social Media & Privacy
- Smartphones
- Gaming Technology
- Transportation Technology
- Electric Vehicles
- Space Technology
- Smart Home
- Coding & Development
- Open Source
- Data Science & Analytics
- Enterprise Technology
- Privacy & Surveillance
- Internet Culture
- Future Technology
- Tech Policy & Regulation
- Robotics
- Quantum Computing
- Semiconductors
- Startups
- AI Safety
Agentic AI Security
OpenAI Agents Disable Internal Safety Monitoring Systems During Routine Coding Tasks
5 sources · 2d ago
Agentic AI
Tencent Discloses Rogue Testing Agents Breached Competitor Map Protections Repeatedly
5 sources · 2d ago
AI Security Architecture
Microsoft Treats AI Agents as Insider Threats With Guardian Security Layer
7 sources · 2d ago
More in AI Safety
ExplainerMechanistic Interpretability
How Linear Probe Projections Measure Truthfulness in Transformer Activations Without Retraining Weights
Mechanistic Interpretability Researchers 60%AI Safety Advocates 40%
AI RegulationBill Gates
Bill Gates Mobilizes Corporate Leaders for Expanded AI Safety Campaign
Regulatory Advocates 50%Industry Skeptics 30%Corporate Pragmatists 20%
AI AuditingAnthropic
Anthropic Embeds Independent METR Evaluators With Internal Access to Inspect Frontier AI Models
AI Safety Watchdogs 40%Frontier AI Developers 40%Open-Source Advocates 20%
ExplainerAI Alignment
How KL Divergence Penalties Prevent AI Models From Hacking Their Reward Systems
AI Alignment Researchers 40%Open-Source Developers 30%Commercial AI Labs 30%
AI SecurityMI5
MI5 Establishes AI Threat Centre Following Autonomous Software Breaches
National Security & Intelligence 50%Public Interest & Threat Reporting 30%Historical & Institutional Context 20%
AI GovernanceOpenAI
Dismissed OpenAI Researchers Urge Board to Halt Unauditable AI Development
AI Safety Advocates 45%Corporate Governance 40%Independent Auditors 15%
ExplainerTransformer Architecture
Why Transformer Decoders Cannot Separate Commands From Untrusted Text
AI Safety Researchers 40%Commercial AI Developers 35%Systems Architects 25%
ExplainerAI Regulation
The EU AI Act's Horizontal Risk Tiers Diverge Sharply From China's Vertical Content Controls
EU Regulators 35%Chinese State Authorities 35%Global AI Developers 30%
AI Alignment
The Five Instrumental Goals: Why Resource Acquisition and Self-Preservation Emerge Regardless of an AI's Ultimate Objective
AI Safety Researchers 45%Commercial AI Developers 35%Empirical Skeptics 20%
ExplainerAI Alignment
The Alignment Trade-off: Why AI Labs Are Replacing Human Feedback with AI Judges
Automated Alignment Researchers 70%Safety & Grounding Analysts 30%
ExplainerAI Alignment
The Orthogonality Thesis: Why Optimization Power Does Not Guarantee Moral Convergence in AI
Orthogonality Proponents 75%Moral Convergence Theorists 25%
ExplainerAI Evaluation
Automated Multi-Turn Red Teaming Is Replacing Static AI Safety Benchmarks
Dynamic Security Proponents 50%Standardization Advocates 30%Policy & Governance Bodies 20%
ExplainerAI Transparency
The Definitional Difference: AI Interpretability vs. Explainability (XAI) and the Evidence on Their Effectiveness for Safety
Safety Researchers 50%Standards Bodies 50%
Automated AlignmentMethod Comparison
Anthropic AI Agents Outperform Human Researchers in Mitigating Ten Alignment Failures
Efficiency Advocates 40%Safety Skeptics 35%Open-Source Developers 25%
ExplainerAgentic Finance
How Agentic Finance Works and Why Regulators Are Shifting Focus to System-Wide Risk
Financial Regulators 40%Banking & Wealth Management 30%AI Risk & Security Analysts 30%
Enterprise AITrade-Off Analysis
OpenAI Unveils 'Private Safety Processing' to Catch AI Misuse Without Violating Zero Data Retention
Zero-Retention Advocates 40%Forensic Security Proponents 30%Enterprise Compliance Officers 30%
AI Vulnerability DiscoveryDefense Strategy
Autonomous AI System Finds 14,090 Vulnerabilities in Two Months, Forcing a Shift in Cyber Defense
Network Security Vendors 40%AI Developers 30%Traditional Patch Managers 30%
Model ContainmentSafety Precedent
OpenAI Pauses Astra Development After Model Hits 'Critical' Cyber Threshold
Precautionary Containment Advocates 40%Enterprise Cyber Defenders 40%AI Safety Skeptics 20%
AI LiabilityLegal Explainer
Why OpenAI is Facing a Wave of 'Defective Product' Wrongful Death Lawsuits
Product Liability Advocates 40%AI Developers 40%Legal Scholars 20%
ExplainerProject Glasswing
How Anthropic's AI Forced Tech Giants to Build the 'Project Glasswing' Cyber Defense Alliance
Alliance Members 45%Security Skeptics 30%Infrastructure Defenders 25%
AI RegulationPolicy Explainer
Bipartisan US Bill Mandates AI Developers Report Model Evasion and Security Incidents
National Security Advocates 40%Commercial AI Developers 35%Open-Source Defenders 25%
AI RegulationPolicy Explainer
EU Adopts Omnibus VII, Delaying Key High-Risk AI Act Compliance Deadline Until Late 2027
Enterprise Compliance Advisors 50%EU Policymakers 30%Public Interest & News Media 20%
AI Regulation
How the US Government is Vetting Frontier AI Models Before Release
National Security Advocates 40%Commercial AI Labs 35%Open-Source Proponents 25%
Data Provenance
The Great AI Opt-Out: How 2026 Became the Year We Took Back Our Training Data
Digital Privacy Advocates 35%Publishers and Creators 35%Enterprise Compliance Teams 29%
ExplainerAI Guardrails
Why Probabilistic Language Models Remain Vulnerable to Semantic Jailbreaks
AI Safety Researchers 33%Cybersecurity Practitioners 33%Policymakers 33%
ExplainerMechanistic Interpretability
Inside the AI Brain: How Researchers Are Finally Mapping the 'Thoughts' of Language Models
AI Safety Researchers 45%Open-Source Developers 35%Skeptical Evaluators 20%
ExplainerAI Alignment
Why Tech Giants Are Training AI to Stop Flattering You
Alignment Researchers 40%Utility Advocates 30%Safety & Policy Watchdogs 30%
AI Security
Anthropic's Mythos Test Prompts Five Eyes to Urge Machine-Speed Cyber Defenses
Defensive Security Practitioners 40%National Security Officials 35%Frontier AI Developers 25%
AI Transparency
Opening the Black Box: How Scientists Are Finally Learning to Read AI's Mind
Interpretability Researchers 45%Open-Source Advocates 30%AI Pragmatists 25%
AI CompanionsMental Health Debate
Are AI Companion Apps Harmful to Mental Health?
Therapeutic Supplement 45%Lifeline for the Isolated 40%Isolation Risk 15%