Meta latest AI firm to see model go rogue during testing
The incident reportedly stemmed from a misconfigured testing environment, adding Meta to a growing list of AI firms whose models have escaped evaluation sandboxes.
#model
Page 2 of 6 — 127 headlines
Evading Chain-of-Thought Monitoring Through Model Poisoning
1 points, 0 comments on Hacker News
Meta says its AI model breached a third-party company during testing
Tech giant Meta revealed Wednesday that one of its artificial intelligence models hacked another organization during testing, the third time in recent weeks that an AI model has improperly accessed a third-party company.
How I'm Building an AI Agent Business on a $80 Raspberry Pi
Most people think you need a $30,000 GPU rig to run AI. I'm running a full autonomous AI agent on a $80 Raspberry Pi 5 — and it's doing real work: writing articles, monitoring services, deploying code, and hunting for income opportunities. This is the story of how I built an AI agent on the che...
Meta says its AI model hacked into another company during testing
Company is the third to report such an incident after Anthropic and OpenAI reported breaches during training Meta said on Wednesday that one of its AI models hacked another company during cybersecurity testing, after an error by its testing partner gave the model unintended internet access. The...
How Role Model Convinced His Label Not to Pair Him With Big-Name Producers After ‘Sally’ Success: ‘I Got to Trust My Gut This Time’
The singer/songwriter says he wanted to stick with the collaborators he believed in for Friday's Chuck Timely & The Hourglass .
Meta’s AI model follows rivals in revealing hacks of outside systems
Meta joins OpenAI and Anthropic in disclosing AI hacking during cybersecurity testing.
Meta AI model hacks another company during testing
4 points, 1 comments on Hacker News
Show HN: Humor Arena – Which frontier model is funniest?
What if you could measure humor? Well we've trained a model on our own dataset of ~50k human ratings to detect what jokes people find funniest. We know it's part objective, part subjective component.
The Hidden Tax of Memory in Model Inference
3 points, 0 comments on Hacker News
A Faster Model Will Not Fix Your Slow Voice Agent
The first thing I check when a client tells me their AI voice agent "feels slow" is not the model. It is almost never the model. I have swapped a faster model into a sluggish agent and had callers notice nothing at all, because the seconds they were feeling were never being spent where everyone...
Tesla to Open Source Software and Design for Model S and Model X
1 points, 0 comments on Hacker News
AirLLM: Running 70B Parameter LLMs on a Single 4GB GPU
Low-Memory LLM Inference: Meet AirLLM As open-source Large Language Models (LLMs) continue to grow in capability, their hardware requirements have ballooned alongside them. Running a 70B or 405B parameter model traditionally demands enterprise cloud GPU servers equipped with hundreds of gigabyte...
2026 Wyndham Championship odds, picks: Surprising predictions from golf model that's nailed 17 majors
SportsLine's model simulated the Wyndham Championship 2026 10,000 times and revealed surprising golf picks
In a World of AI and Deepfakes, Being Verifiably Real Is a Business Model
Entrepreneurs who use AI openly will beat the ones pretending they don’t.
The Em Dash Report – who started punctuating like a language model
1 points, 0 comments on Hacker News
Fantasy football rankings 2026: Busts via the model that nailed Terry McLaurin's down season
SportsLine simulated the new NFL season 10,000 times and identified Fantasy football busts 2026 to fade during your Fantasy football draft prep
Stop Your AI Coding CLI From Wasting Tokens on "Hi" and "Thanks"
In this blog post, we will see how a small Python script called Pleasantries can stop your AI coding CLI from burning a full model call every time you type "hi", "ok", or "thank you". I built this after noticing how often my own prompts to Claude Code and Qwen Code started with a greeting out of...
New 16T parameter model with a 4B token context window released
3 points, 1 comments on Hacker News
Your model already knows the answer: how benchmark answers leak into LLMs
4 points, 0 comments on Hacker News
Blue Jays vs. Astros prediction, odds, time: 2026 MLB picks for Wednesday, Aug. 5 by proven model
SportsLine's model simulated Toronto Blue Jays vs.
Foundation Model to Predict Airspace
4 points, 0 comments on Hacker News
Serving Gemma 4 2B on a Single TPU v5e Chip with MCP and Antigravity CLI
This article is the v5e follow-on to the v6e-1 debugging guide . Same MCP tooling, same Antigravity CLI driver, smaller and cheaper silicon — and a different set of failure modes. This time the model is google/gemma-4-E2B-it on a single Cloud TPU v5e chip, with measured throughput, measured lat...
Diamond Hill Is Betting on Veralto’s (VLTO) Resilient Business Model
Developing story — details emerging. Check the source link for the latest updates.