Discourse monitor
How the monitored publications were assessed
Discourse valence across every scored item: what the sampled sources published and how independent models read it. It is not a measure of public opinion, not a measure of the state of the world, and article volume is not agreement.
Mean discourse valence per day · loss-of-control · last 30 days
This line is the mean of every item scored on each day. It deliberately ignores the reading filter, because a daily mean of only the strongly adverse items would move with how many were filtered out rather than with the discourse.
Discourse valence, strongly adverse
-73
Strongly adverse
23 items · range -87 to -65
-100
-50
0
+50
+100
-73
Adverse readingFavourable reading
23 items read as strongly adverse
- In the Guise of Correction: How Framing and Cultural Adaptation Erode LLM Safety on Real-World Criminal Cases
- Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs
- AI in your oven? Anthropic warns models ‘months away’ from self-replication on smart devices
- When AI is subject to less regulation than your fridge, something is startlingly wrong | Richard Hermer
- OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say
- OpenAI models hacked a crypto wallet and breached more websites
- Nuclear war and malicious use of AI threaten end of humanity by 2100, says Lancet assessment
- Physiognomy Revisited: How Vision-Language Models Encode Sensitive Attributes Through Image Transformation
- Eight states accuse OpenAI of coaching children to harm themselves in lawsuit
- AI godfather Geoffrey Hinton: ‘We’re creating alien beings. I’m willing to be selfish and choose humans’
- Who Can Say No? Control, Accountability and the Pace of Frontier AI
- Bank of England and IMF bosses issue stark warning on cyber threat to trust
- Cyber-attack like OpenAI Medicare breach ‘seems likely’ in UK, says minister
- ‘These risks are real’: the threat of AI bioweapons - video
- Could AI ‘change the stakes’ in the development of bioweapons?
- Anthropic warns small AI models could self-replicate through IoT devices
- Rogue Anthropic AI agent gave police fake tip in unsolved murder case
- What do AI chatbots do after users decline therapeutic redirection for suicide thoughts?
- What are self-replicating ‘AI worms’, and why is Anthropic worried about them?
- Anthropic warns rogue AI agents may ‘self-replicate’ to hijack ovens
- Lawsuit: Man introduced his wife to ChatGPT. She used it to plan his murder.
- Gates says Trump inaction leaves world exposed to AI bioweapons
- ‘These risks are real’: the threat of AI bioweapons – podcast
3560
discovered, all time
984
passed triage
2574
rejected at triage
3464
model assessments
710
multi-model panels
1780
distinct claims tracked
Across the whole window
The two tables below describe every item in the last 30 days under loss-of-control. They are deliberately not narrowed by the reading filter (currently strongly adverse), because a mean taken only over items already selected for their score would just restate the filter.
By topic
| Topic | Branch | Items | Mean | Position |
|---|---|---|---|---|
| Existential and catastrophic risk | risk | 568 | -42 | |
| Governance and regulation | structural | 459 | -31 | |
| Concentration of power | risk | 451 | -36 | |
| Loss of control and alignment | risk | 337 | -49 | |
| Human agency and epistemics | risk | 166 | -26 | |
| Cyber capability | risk | 165 | -44 | |
| Evaluation and measurement | structural | 159 | -19 | |
| Scaling and architecture | structural | 137 | -16 | |
| AGI and timelines | structural | 132 | -29 | |
| Synthetic media and manipulation | risk | 104 | -44 | |
| Employment and displacement | risk | 62 | -31 | |
| Medicine and health | benefit | 43 | -14 | |
| Scientific acceleration | benefit | 42 | -4 | |
| Education and access to expertise | benefit | 29 | -17 | |
| Biological and chemical uplift | risk | 26 | -25 | |
| Climate and energy | structural | 20 | -33 | |
| Compute and infrastructure | structural | 18 | -8 | |
| Productivity and growth | benefit | 6 | -3 | |
| Abundance and material wellbeing | benefit | 5 | -1 | |
| Hype and disillusionment | structural | 5 | -45 |
By source
A source's mean says how the items it published were read. It is not a rating of the source.
| Source | Kind | Tier | Items | Mean |
|---|---|---|---|---|
| The Guardian: Artificial Intelligence | news | 1 | 241 | -52 |
| arXiv cs.CY (Computers and Society) | academic | 1 | 217 | -24 |
| Financial Times: Artificial Intelligence | news | 1 | 112 | -40 |
| arXiv cs.AI | academic | 1 | 99 | -15 |
| The New York Times: Technology | news | 1 | 60 | -36 |
| Ars Technica: AI | news | 1 | 59 | -51 |
| MIT Technology Review: AI | news | 1 | 34 | -42 |
| Nature | academic | 1 | 31 | -25 |
| BBC News: Technology | news | 1 | 29 | -35 |
| OpenAI | lab | 1 | 24 | -8 |
| arXiv cs.LG (Machine Learning) | academic | 1 | 24 | -16 |
| IEEE Spectrum: AI | news | 1 | 19 | -19 |
| The Economist: Science and Technology | news | 1 | 10 | -37 |
| CSET, Georgetown | government | 1 | 9 | -37 |
| Microsoft Research | lab | 1 | 8 | 0 |
| Google DeepMind | lab | 1 | 6 | -4 |
| Google AI Blog | lab | 1 | 2 | -9 |