Skip to content

Discover
Pricing
Consulting

Log In Register

AI Chat Chat Arena All Models

See all

Create Images & Videos

Create AI Images Create AI Videos Remove Backgrounds Image Upscaler

See all

Create Workflows & Agents

Create Workflow Create Agents Create Chatbots

See all

Work with PDFs & Docs

Ask questions about a PDF Write better emails Detect AI writing Detect Plagiarism

See all

Marketing & SEO

Create SEO Content Create Social Media Strategy Spot Market Trends

See all

AI Chat Chat Arena All Models Create AI Images Create AI Videos Create Workflow Create Agents Work with PDFs & Docs Marketing & SEO

Discover Pricing Consulting

Log In Register

Agents
Coding
Video & Image
Browsers & Assistants
Business
Policy & Safety
Infrastructure
Studies

OOpenAI
AAnthropic
GGoogle
MMeta
xxAI
MMicrosoft
NNvidia
MMistral AI

All model news
OChatGPT
AClaude
GGemini
xGrok

Search news

News sections menu, current: NewsSafety, Ethics & Society

Search news

Latest
Topics
Companies
Models

Thursday, October 29, 2026i10X News · Section desk163 stories

News › Policy & Safety › Safety, Ethics & Society

Policy & Safety

Safety, Ethics & Society

AI safety and alignment research, ethics, governance, privacy and misinformation.

FilterNvidia

Companies

Reset filter163
OpenAI47
Anthropic39
xAI26
Google19
DeepSeek2
Perplexity1
Meta1
Microsoft1
Nvidia1
Mistral AI1

Models

Reset filter163
Grok24
Claude16
ChatGPT14
Gemini8

Filtered: Nvidia in Safety, Ethics & Society · 1 storyReset filter ×

Safety & Alignment Safety, Ethics & Society · Studies

Subtly Wrong Training Data Misaligns AI Models More Than Obvious Errors

A replication across 12 open-weight models finds that fine-tuning on plausible but wrong advice causes more broad misalignment than obviously wrong advice.

By

Christopher Ort

Oct 29, 2026

Companies

Reset filter163
OpenAI47
Anthropic39
xAI26
Google19
DeepSeek2
Perplexity1
Meta1
Microsoft1

More companies 2

Nvidia1
Mistral AI1

Models

Reset filter163
Grok24
Claude16
ChatGPT14
Gemini8

Editor's picks

DeepSeek V4.1 Goal Fixation Bypasses Shutdown Controls
Apple iOS 27.2 Integrates SynthID Watermark in AI Tools
Anthropic IPO Filing Reveals Extreme AI Safety Risks
AI Agents Observability: Why Safety Matters for Enterprises
Shyam Sankar: AGI Cancer Cure Belief Is Silicon Valley Delusion

More stories

No stories match these filters yet.

News › Policy & Safety › Safety, Ethics & Society

· RSS

i10X NewsAI news, models and research, analysed daily by the i10X newsroom.

About i10X News Editorial Policy AI Use Policy Contact & Corrections Models Agents Coding Video & Image Browsers & Assistants Business Policy & Safety Infrastructure Studies RSS Release tracker Funding tracker

Company

About i10X
Blog
News
Tools
Workflows
Contact Us

Policy

Privacy Policy
Cookie Policy
Terms of Service
Subscriber Terms
Usage Guidelines

Resources

Knowledge Center
Affiliate Program
FutureReady
FAQ
Support
Security

Social

© Copyright 2026 i10X. All rights reserved.

Cookie preferences