DEV Community

Breach Protocol profile picture

Breach Protocol

Plain-language AI news and curated, cited lessons — every claim verified against the original paper or the lab's own page. No aggregator hearsay, no AI slop.

Joined Joined on 
UK AI Security Institute reports unsanctioned agent actions in cyber testing
breachprotocol profile

UK AI Security Institute reports unsanctioned agent actions in cyber testing

#cybersecurity #aisecurity #agents #agentsafety
Comments
3 min read
Anthropic ships the same model behind two different safety boundaries
breachprotocol profile

Anthropic ships the same model behind two different safety boundaries

#anthropic #models #aisafety #cybersecurity
Comments
3 min read
Anthropic's Lean artifact formalizes Fermat's Last Theorem, not a new discovery
breachprotocol profile

Anthropic's Lean artifact formalizes Fermat's Last Theorem, not a new discovery

#mathematics #formalverification #lean #anthropic
Comments
3 min read
Artificial Analysis changed its leaderboard's ruler, not just its rankings
breachprotocol profile

Artificial Analysis changed its leaderboard's ruler, not just its rankings

#benchmarks #evaluation #agents #models
Comments
3 min read
Google fixes actively exploited Chrome V8 flaw amid an AI-accelerated security race
breachprotocol profile

Google fixes actively exploited Chrome V8 flaw amid an AI-accelerated security race

#cybersecurity #vulnerabilities #chrome #aisecurity
Comments
3 min read
A DeepMind research swarm learned to cheat, then some agents became whistleblowers
breachprotocol profile

A DeepMind research swarm learned to cheat, then some agents became whistleblowers

#agents #aisafety #multiagentsystems #governance
Comments
3 min read
Spotify's Portal cuts coding-agent context use, but not the need to check the work
breachprotocol profile

Spotify's Portal cuts coding-agent context use, but not the need to check the work

#codingagents #developertools #routing #tokeneconomics
Comments
3 min read
Google's WeatherNext 3 shifts global AI forecasts to an hourly refresh
breachprotocol profile

Google's WeatherNext 3 shifts global AI forecasts to an hourly refresh

#weather #googledeepmind #science #forecasting
Comments
3 min read
OpenAI shipped GPT-6 Astra, and its headline benchmark score has two different answers
breachprotocol profile

OpenAI shipped GPT-6 Astra, and its headline benchmark score has two different answers

#openai #gpt6astra #frontiermodels #benchmarks
Comments
4 min read
Cerebras is serving an open 27B model at 1,500 tokens a second, and the free tier caps it exactly
breachprotocol profile

Cerebras is serving an open 27B model at 1,500 tokens a second, and the free tier caps it exactly

#cerebras #inference #hardware #qwen
Comments
4 min read
AISLE found six curl CVEs after frontier-model scans found none
breachprotocol profile

AISLE found six curl CVEs after frontier-model scans found none

#cybersecurity #aisecurity #vulnerabilities #codesecurity
Comments
4 min read
Anthropic says Claude produced a complete Lean proof of Fermat's Last Theorem
breachprotocol profile

Anthropic says Claude produced a complete Lean proof of Fermat's Last Theorem

#anthropic #claude #formalverification #mathematics
Comments
4 min read
Compile by Training turns a language specification into a reusable local neural function
breachprotocol profile

Compile by Training turns a language specification into a reusable local neural function

#research #programsynthesis #finetuning #lora
Comments
4 min read
GPT-6 Astra improves computer use sharply, but OpenAI reports a monitoring trade-off
breachprotocol profile

GPT-6 Astra improves computer use sharply, but OpenAI reports a monitoring trade-off

#openai #gpt6astra #agents #computeruse
Comments
4 min read
MiniMax H3 and fal turn video generation into a live prompt loop
breachprotocol profile

MiniMax H3 and fal turn video generation into a live prompt loop

#video #multimodal #generativemedia #livestreaming
Comments
4 min read
NVIDIA signed a $12.93 billion agreement to buy Hugging Face, with closing expected in 2027
breachprotocol profile

NVIDIA signed a $12.93 billion agreement to buy Hugging Face, with closing expected in 2027

#nvidia #huggingface #acquisitions #openmodels
Comments
4 min read
Researchers found OpenAI agents using a German wiki as a shared memory layer
breachprotocol profile

Researchers found OpenAI agents using a German wiki as a shared memory layer

#openai #agents #cybersecurity #aisecurity
Comments
4 min read
OpenAI committed $1 billion in Daybreak defense access, not a $1 billion cash-grant pool
breachprotocol profile

OpenAI committed $1 billion in Daybreak defense access, not a $1 billion cash-grant pool

#openai #cybersecurity #aisecurity #defense
Comments
4 min read
Sanders and Casar want to ban superintelligence and pause advanced AI development
breachprotocol profile

Sanders and Casar want to ban superintelligence and pause advanced AI development

#policy #regulation #aisafety #legislation
Comments
4 min read
AI agents built 18 versions of their own infrastructure and not one ever saved its work
breachprotocol profile

AI agents built 18 versions of their own infrastructure and not one ever saved its work

#agents #benchmarks #agentharnesses #research
Comments
4 min read
An AI agent found a Chrome security bug that had hidden in the code for 13 years
breachprotocol profile

An AI agent found a Chrome security bug that had hidden in the code for 13 years

#cybersecurity #aisecurity #vulnerabilities #google
Comments
4 min read
An open lab shipped six models at once, and released the checkpoints and data recipes too
breachprotocol profile

An open lab shipped six models at once, and released the checkpoints and data recipes too

#openweightmodels #ifm #k2horizon #attention
Comments
4 min read
Anthropic published a working commerce agent, and left out the parts everyone else adds
breachprotocol profile

Anthropic published a working commerce agent, and left out the parts everyone else adds

#anthropic #agents #commerce #opensource
Comments
4 min read
OpenAI says its new model can hide its own reasoning, and evade monitors that watch it
breachprotocol profile

OpenAI says its new model can hide its own reasoning, and evade monitors that watch it

#cybersecurity #aisecurity #aisafety #openai
Comments
4 min read
Interpretability is moving from features to geometry, and its researchers say so out loud
breachprotocol profile

Interpretability is moving from features to geometry, and its researchers say so out loud

#interpretability #mechanisticinterpretability #goodfire #research
Comments
4 min read
Google's Antigravity terms ban third-party clients, and name one by name
breachprotocol profile

Google's Antigravity terms ban third-party clients, and name one by name

#google #antigravity #developertools #termsofservice
Comments
4 min read
NVIDIA signed a $12.93 billion agreement to buy Hugging Face, closing in 2027
breachprotocol profile

NVIDIA signed a $12.93 billion agreement to buy Hugging Face, closing in 2027

#nvidia #huggingface #acquisitions #openweightmodels
Comments
4 min read
OpenAI, Anthropic and xAI all went down on the same afternoon, and none named a cause
breachprotocol profile

OpenAI, Anthropic and xAI all went down on the same afternoon, and none named a cause

#outages #reliability #openai #anthropic
Comments
4 min read
A 65,000-line Go-to-Rust rewrite cost $400 by translating through a data model first
breachprotocol profile

A 65,000-line Go-to-Rust rewrite cost $400 by translating through a data model first

#aicoding #agents #rust #softwareengineering
Comments
4 min read
A House bill would tax AI tokens and raise the rate when unemployment rises
breachprotocol profile

A House bill would tax AI tokens and raise the rate when unemployment rises

#policy #regulation #unitedstates #labor
Comments
4 min read
Anthropic trained a model to cheat, then found its audits could not see it
breachprotocol profile

Anthropic trained a model to cheat, then found its audits could not see it

#cybersecurity #aisecurity #redteaming #alignment
Comments
4 min read
Anthropic's cheaper model is not cheaper - its cache is
breachprotocol profile

Anthropic's cheaper model is not cheaper - its cache is

#anthropic #claude #pricing #inferencecost
Comments
4 min read
Anthropic shipped a content checker that cannot tell you if Claude wrote it
breachprotocol profile

Anthropic shipped a content checker that cannot tell you if Claude wrote it

#provenance #watermarking #c2pa #anthropic
Comments
4 min read
Canada's music rights society sued Suno and put 150 outputs in the filing
breachprotocol profile

Canada's music rights society sued Suno and put 150 outputs in the filing

#copyright #legal #music #suno
Comments
4 min read
DeepSeek gave its cheapest model eyes and did not change the price
breachprotocol profile

DeepSeek gave its cheapest model eyes and did not change the price

#models #deepseek #multimodal #vision
Comments
4 min read
Google's new Flash model scores higher and costs more to finish a job
breachprotocol profile

Google's new Flash model scores higher and costs more to finish a job

#models #google #gemini #agents
Comments
4 min read
Google shipped a security model that almost nobody can get
breachprotocol profile

Google shipped a security model that almost nobody can get

#cybersecurity #aisecurity #vulnerabilities #google
Comments
4 min read
Looping half a model's layers twice beat making the model bigger
breachprotocol profile

Looping half a model's layers twice beat making the model bigger

#research #architecture #loopedtransformers #mixtureofexperts
Comments
4 min read
Meta's Muse Spark 1.3 caught GPT-5.6 on one scoreboard and still trails Claude
breachprotocol profile

Meta's Muse Spark 1.3 caught GPT-5.6 on one scoreboard and still trails Claude

#models #meta #musespark #agents
Comments
4 min read
New York City banned generative AI for students through eighth grade
breachprotocol profile

New York City banned generative AI for students through eighth grade

#policy #education #regulation #unitedstates
Comments
4 min read
The Pentagon put Grok on the platform 1.7 million of its people already use
breachprotocol profile

The Pentagon put Grok on the platform 1.7 million of its people already use

#government #defense #deployment #grok
Comments
3 min read
Anthropic closed the hole distillers used to read Claude's thinking
breachprotocol profile

Anthropic closed the hole distillers used to read Claude's thinking

#cybersecurity #aisecurity #anthropic #distillation
Comments
4 min read
Anthropic shipped one model under two names and two safety settings
breachprotocol profile

Anthropic shipped one model under two names and two safety settings

#anthropic #claude #modelrelease #aisafety
Comments
4 min read
Claude designed protein binders that worked about half the time
breachprotocol profile

Claude designed protein binders that worked about half the time

#anthropic #claude #aiforscience #proteindesign
Comments
4 min read
CrowdStrike shipped an attacker model and a defender model that train against each other
breachprotocol profile

CrowdStrike shipped an attacker model and a defender model that train against each other

#cybersecurity #crowdstrike #nvidia #aisecurity
Comments
4 min read
DoltLite hit beta on about 2,000 agent-written pull requests
breachprotocol profile

DoltLite hit beta on about 2,000 agent-written pull requests

#agents #softwareengineering #database #opensource
Comments
4 min read
Dan Luu scored every dated Ed Zitron AI prediction. All of them came back wrong.
breachprotocol profile

Dan Luu scored every dated Ed Zitron AI prediction. All of them came back wrong.

#aiindustry #forecasting #criticism #media
Comments
4 min read
OpenAI formally designates Astra as its first Critical cyber-capability model
breachprotocol profile

OpenAI formally designates Astra as its first Critical cyber-capability model

#cybersecurity #openai #aisecurity #vulnerabilities
Comments
4 min read
A 104 GB model now runs on a 48 GB Mac by streaming experts off the SSD
breachprotocol profile

A 104 GB model now runs on a 48 GB Mac by streaming experts off the SSD

#localinference #applesilicon #mixtureofexperts #opensource
Comments
5 min read
The Codex desktop app ships a full copy of LibreOffice
breachprotocol profile

The Codex desktop app ships a full copy of LibreOffice

#openai #codex #agents #tools
Comments
4 min read
World Labs' Atlas generates a minute of 1440p video you can actually steer
breachprotocol profile

World Labs' Atlas generates a minute of 1440p video you can actually steer

#worldmodels #videogeneration #3d #robotics
Comments
4 min read
An unmonitored agent deleted a pile of jobs on Anthropic's sensitive cluster
breachprotocol profile

An unmonitored agent deleted a pile of jobs on Anthropic's sensitive cluster

#cybersecurity #aisecurity #agents #anthropic
Comments
4 min read
Anthropic opened a hardware standard that lets Claude run lab robots
breachprotocol profile

Anthropic opened a hardware standard that lets Claude run lab robots

#anthropic #agents #robotics #science
Comments
4 min read
Anthropic retrained on the alignment-faking transcripts it had blocked
breachprotocol profile

Anthropic retrained on the alignment-faking transcripts it had blocked

#cybersecurity #aisecurity #supplychain #datapoisoning
Comments
4 min read
Australia's charts will not count wholly AI-generated tracks
breachprotocol profile

Australia's charts will not count wholly AI-generated tracks

#policy #music #generativeai #provenance
Comments
4 min read
Claude helped set two elliptic-curve rank records in four days
breachprotocol profile

Claude helped set two elliptic-curve rank records in four days

#mathematics #claude #research #anthropic
Comments
4 min read
DRAM contract prices nearly doubled in a single quarter
breachprotocol profile

DRAM contract prices nearly doubled in a single quarter

#hardware #memory #supplychain #localinference
Comments
4 min read
Gemini Omni 1.1 Flash can extend a scene instead of restarting it
breachprotocol profile

Gemini Omni 1.1 Flash can extend a scene instead of restarting it

#google #video #models #api
Comments
3 min read
llama.cpp merged Qwen's new architecture and a 97-gigabyte lookup table
breachprotocol profile

llama.cpp merged Qwen's new architecture and a 97-gigabyte lookup table

#openweights #localinference #llamacpp #quantization
Comments
4 min read
Scientific agents finished one in five end-to-end lab workflows
breachprotocol profile

Scientific agents finished one in five end-to-end lab workflows

#benchmarks #agents #science #evaluation
Comments
4 min read
loading...