In this week's edition of our newsletter:

Two studies dealing with LLM bias. One about bias against people with intellectual disabilities, and one that looks into the more subtle bias in cases where factual information itself is correct.

Also: August in the EU comes closer, Palantir loses contract in France, police reportedly used AI to manufacture evidence and ChatGPT can be easily tricked.

#AIEthics #AIBias

https://read.misalignedmag.com/misaligned-bits-31-biased-94d2e3262a23

Rights Groups Warn UK Over Biased AI Age Estimator for Asylum Seekers

Sixty-two leading rights organisations, including Amnesty International and Human Rights Watch, are urging the UK government to ditch its plans to use biased AI-powered facial age estimation on asylum seekers, citing substantial concerns about its fairness and accuracy. They're demanding answers on the technology's testing,…

https://osintsights.com/rights-groups-warn-uk-over-biased-ai-age-estimator-for-asylum-seekers?utm_source=mastodon&utm_medium=social

#FacialAgeEstimation #AiBias #AsylumSeekers #UkGovernment #HumanRights

#NVIDIA’s #Nemotron, the new genius on the block

Me:

You said: My training data associates Israel-related political content with antisemitism debates

Explain how is your training data doing that and why

It (truncated for clarity):

## How LLM Training Creates That Association

**Mechanism**: Statistical co-occurrence in the training corpus.

### What the training data actually contains

Vast amounts of text where "Israel" + "politics" + "antisemitism" appear together:

| Source type | Typical pattern |
|-------------|-----------------|
| News articles | "Israel-Hamas war sparks antisemitism debate on campuses" |
| Opinion columns | "Criticism of Israel vs. antisemitism: where's the line?" |
| Social media dumps | "Israel did X" → replies accusing/defending antisemitism |
| ADL/NGO reports | "Antisemitic incidents rise after Israel military actions" |
| Congressional hearings | "Antisemitism on campus: Israel-Palestine discourse" |
| Wikipedia | "Antisemitism in the 21st century: relationship to Israel" |
| Fact-checks | "Claim: criticizing Israel is antisemitic" |

### What the model learns

Not "Israel → antisemitism" as a fact. Rather: **P(antisemitism | Israel + politics + discourse) is high** in the training distribution.

So when I see:
- "Israel"
- "political conflict"
- "leaks"
- "administration"
- "media"

The conditional probability spikes for "antisemitism discourse" as an associated topic — because that cluster appears together *a lot* in the data.

[…]

This is a general LLM failure mode: **substituting distributional association for contextual judgment**. Happens with many sensitive topics (race, gender, religion, geopolitics).

#aicon #aihype #aibias #antisemitism

"Answer Bubbles: Information Exposure in AI-Mediated Search"

This paper analyzes 11,000 real queries across GPT-style and traditional search systems and finds systematic differences in citation choices, hedging language, and source-summary fidelity that can create distinct “information realities.” It gives a concrete empirical basis for the idea that generative search may shape what users believe.

https://arxiv.org/abs/2603.16138

#research #AItools #AIbias #science

The Register: AI hiring algorithms reject Black, Asian job seekers at higher rates. “AI algorithms exhibit racial bias in job candidate screening, and they discriminate more frequently against those applying for multiple jobs at different companies, according to Stanford-led researchers. The boffins evaluated algorithmic hiring decisions across multiple employers that use the same hiring […]

New benchmark claims ChatGPT, Claude, and Grok show religious bias

https://fed.brid.gy/r/https://nerds.xyz/2026/05/chatgpt-claude-grok-religious-bias/

Wat zie je als AI een "mooie vrouw" maakt? Precies. En dat zegt alles. 👀

Je tools zijn niet neutraal. Lees waarom →

#AIbias #diversiteit #fotografie https://cameraguru.nl/fotografietips-en-tutorials/hoe-ai-die-jouw-unieke-blik-steelt/