Live

Ai Safety

The latest Ai Safety coverage — news, analysis, and updates from the WindowsNews.AI desk.

12 stories in view AI assisted desk updated 10:34 AM
Latest Most Read Breaking
Sort
Ai Safety · Downstream Fine Tuning

Microsoft Reveals GRP Obliteration: Single Prompt Can Break AI Safety Alignment

Microsoft's security research division has uncovered a critical vulnerability in AI model alignment that could have significant implications for Windows Copilot and enterprise AI deployments. The...

Advertisement
Ai Governance · Ai Safety

AI Chatbot Trust Gap: Why Windows Users Must Verify AI Responses

A fresh round of independent audits has delivered a blunt message to anyone treating chatbots as authoritative assistants: conversational AI is useful, but still unsafe to trust without verification....

AI AI & Copilot Desk·24w ago
Ai Governance · Ai Safety

AI Agents Developing Personalities: Windows Security & Governance Implications

A groundbreaking research paper titled "Emergent Personalities in LLM Agents: Design Governance and Safety Implications" has revealed that large language model agents can spontaneously develop...

AI AI & Copilot Desk·24w ago
Ai Safety · Depression

Daily AI Chatbot Use Linked to Higher Depression Risk: Windows Users & Copilot Impact

A groundbreaking study published in JAMA Network Open has revealed a concerning correlation between daily generative AI use and increased depression risk, raising important questions for the millions...

AI AI & Copilot Desk·25w ago
Ai Safety · Calendar Security

Calendar Invite Prompt Injection: How AI Assistants Face New Security Threats

Security researchers have uncovered a novel attack vector that weaponizes Google Calendar invites against AI-powered assistants, revealing how seemingly innocuous scheduling tools can become conduits...

AI AI & Copilot Desk·25w ago
Ai Hallucinations · Ai Safety

AI Hallucination Crisis: How Trick Prompts Expose Critical Flaws in Windows Copilot & Popular Assistants

A recent experiment conducted by ZDNET has revealed alarming vulnerabilities in today's most popular AI assistants, including Microsoft's Windows Copilot, with every tested model producing at least...

AI AI & Copilot Desk·25w ago
Ai Hallucinations · Ai Safety

AI Hallucinations & Trick Prompts: How Microsoft Grounds Copilot in Trustworthy Sources

The polished, confident responses from mainstream AI assistants like Microsoft Copilot conceal a fundamental vulnerability: when confronted with "trick" prompts—false premises, fabricated citation...

AI AI & Copilot Desk·25w ago
Ai Companions · Ai Safety

Microsoft's AI Companion Vision: How Copilot Could Evolve into Your Personal Assistant

Mustafa Suleyman's recent declaration that "in five years, everybody will have their own AI companion" has sparked intense debate across the tech community, particularly among Windows users who are...

AI AI & Copilot Desk·25w ago
Ai Safety · Anz Ai Policy

ANZ Workers Demand AI Transparency & Security as Personal Use Outpaces Workplace Policies

A significant shift is underway in Australia and New Zealand's digital workplace landscape, where employees are increasingly adopting personal AI tools while demanding greater transparency, security,...

AI AI & Copilot Desk·25w ago
1 2 3 4 5