October 2026, Arbiter featured in Nieman Lab and daily narrative monitoring for newsrooms
From Investigating Stories to Spotting Them Early
Earlier this month, Arbiter was featured in Nieman Lab, leading to over 150 sign-ups from more than 80 organizations across 31 countries.
Since then, we have been working individually with new users to help them get started with Arbiter. These conversations reinforced something important: while Arbiter helps journalists investigate how narratives developed, newsrooms also need to spot emerging stories as conversations unfold. Listening to users has shaped how we have built Arbiter over the past two years, and this past quarter has been no different.
That is why we are introducing daily narrative monitoring in Arbiter. From political claims gaining traction to emerging conversations around technology or public health, journalists can now receive daily summaries of evolving narratives, key actors, and emerging claims, without having to start a new investigation every day. We have been rolling this out to trusted partners over the past few weeks, and we are opening it up more broadly next week.
Follow a Story While It Is Still Forming
Sign up for Arbiter to follow emerging stories across politics, technology, sports, and more.
Recent Updates From SimPPL
At Stanford's Trust & Safety Research Conference, co-founder Swapneel Mehta led a hands-on session on Arbiter and cross-platform, multilingual trust & safety investigations. The session explored using AI to investigate public deception and online harms across social platforms.
Co-founder Dhara Mungra spent two days with the NEST Center for Journalism Innovation and Development, leading a training-of-trainers programme on AI in newsroom workflows. The sessions covered research, editing and visual storytelling, newsroom AI policies, and rapid prototyping.
Stanford Session →NEST Training →
Investigations Conducted Using Arbiter
Anti-LGBTQ Online Hate and Gender Misinformation Trends
Political debates over transgender rights, reports of harassment and disagreements over hate speech shaped online LGBTQ+ discussions in the previous month.
What Arbiter Found
- Conservative commentary accused the political left of organising harassment and using "hate speech" to justify violence. These arguments received more attention than firsthand accounts of harassment.
- Trans users in the UK described increased hostility following the Supreme Court ruling on the legal meaning of "woman", including slurs and misgendering.
- Users responded to hostile content by reporting, blocking and warning others. One newsroom even asked readers to help flag comments its moderators could not keep up with.
Brain-Computer Interface Security Concerns on Social Media
As brain-computer interfaces move towards wider medical use, conversations around them span security vulnerabilities, neural-data privacy and claims of non-consensual brain monitoring.
What Arbiter Found
- Researchers have demonstrated ways BCIs could be manipulated, including through fake brain signals and replayed commands, but the research received limited attention online.
- Claims about remote brain monitoring reached much larger audiences than research into documented BCI vulnerabilities.
- Who controls neural data is also becoming a concern, with discussion around consent, regulation and how neural data should be collected and stored.
- Karotte: An Open-Source Framework for Building Robust RL Environments for Autonomous Agents
An open-source framework by Preference Model for building secure reinforcement learning environments. It provides hardened sandboxes with restricted permissions and process controls to prevent agent privilege escalation and cheating.
- Gemma 4: Google DeepMind's Open Multimodal Model Family
Gemma 4 is Google DeepMind's open-weights model family offering multimodal reasoning across various deployment sizes. It features step-by-step thinking modes, native system prompts, and extended context windows for agentic workflows.
- Headroom: Local Context Compression Layer for AI Agents and LLMs
Headroom is a local context compression tool and proxy designed to lower token usage for AI agents and RAG pipelines. It runs on-device to shrink logs, files, and conversation history while preserving key data via reversible caching.
- Uncivility Rates in Facebook and Instagram Comments
A social media study finds that 4 to 8 percent of general comments on Facebook and Instagram are uncivil. That figure rises to 10 to 15 percent on untrustworthy sources, scaling to hundreds of millions of uncivil comments daily.
- IO-Campaign-Collections: A Curated Archive of Influence Operation Reports
This GitHub repository provides a chronological directory of public threat intelligence reports on covert influence operations. It aggregates analysis on state-sponsored disinformation tactics from firms like Meta and OpenAI.