Educational Testing Service's E-rater algorithm biases essay scores against minority students
The Educational Testing Service's E-rater algorithm, used to grade essays on the GRE and other standardized tests, has been found to systematically give higher scores to students from mainland China and lower scores to African American students compared to human graders. The bias stems from the algorithm's reliance on surface-level metrics like vocabulary and sentence length, which disadvantage certain groups. Despite studies dating back to 1999, the bias persists, and in many states, only a small percentage of essays are reviewed by humans.
- Company involved
- Educational Testing Service
- AI system involved
- E-rater
10 source articles · read the reporting →
Google, Microsoft, and Perplexity AI search results promote racist IQ data
AI-powered search engines from Google, Microsoft, and Perplexity have been surfacing debunked research promoting race science, including false IQ scores for countries. The systems pulled data from a dataset by Richard Lynn, a known proponent of scientific racism. Google removed the offending Overviews after being contacted by WIRED, but the data still appears in featured snippets and other AI tools.
- Company involved
- Google, Microsoft, and Perplexity
- AI system involved
- AI Overviews, Copilot, Perplexity
7 source articles · read the reporting →
NSW Education Standards Authority used AI-generated image in HSC English exam without disclosure
The NSW Education Standards Authority (NESA) used an AI-generated image as a stimulus in the 2024 HSC English exam without disclosing its origin. The image, created by Florian Schroeder using OpenAI's ChatGPT and Dall-E 2, was published on Medium in July 2023. Students suspected AI use due to irregularities in the image, and NESA initially declined to confirm. After the Sydney Morning Herald confirmed the image was AI-generated, NESA stated that students would be marked on their response to the question, not the image's origin.
- Company involved
- NSW Education Standards Authority
- AI system involved
- ChatGPT and Dall-E 2
6 source articles · read the reporting →
Middle schooler beats Edgenuity grading algorithm to get perfect score
A seventh-grade student in the Los Angeles Unified School District received a failing grade on a history assignment graded by Edgenuity's automated scoring algorithm. With help from his mother, a history professor, he reverse-engineered the algorithm by writing a paragraph with relevant keywords and a jumble of words, earning a perfect score. The incident highlights concerns about the accuracy and fairness of automated grading systems in education.
- Company involved
- Los Angeles Unified School District
- AI system involved
- Edgenuity
10 source articles · read the reporting →
Proctortrack data breach exposed student data from online proctoring
Proctortrack, an online proctoring service used by universities, suffered a data breach in September 2020 when its source code was leaked online. An analysis by Consumer Reports found that the code contained hard-coded passwords and exposed the names and email addresses of over 150 students. The company acknowledged the leak but said no harm resulted. Students had been required to use the software, which performed facial recognition and recorded video during exams.
- Company involved
- Proctortrack
- AI system involved
- Proctortrack
10 source articles · read the reporting →
College student uses GPT-3 to create fake blog that reaches #1 on Hacker News
Liam Porr, a college student at UC Berkeley, used OpenAI's GPT-3 language model to generate a fake blog under a fake name. One of the posts reached the number-one spot on Hacker News, fooling tens of thousands of readers. Porr later confessed and retired the blog after two weeks. The experiment demonstrated the ease of creating convincing AI-generated content.
- AI system involved
- GPT-3
10 source articles · read the reporting →
ChatGPT-4 generates false accusations and quotes about law professors
OpenAI's ChatGPT-4 generated false accusations and fabricated quotes about law professors when asked about scandals and crimes involving law professors. The system produced entirely made-up newspaper citations and quotes, falsely alleging misconduct such as harassment and tax fraud. The article reports that these hallucinations could mislead users who trust the generated quotes. The author corrected an earlier error attributing similar results to ChatGPT-4 instead of ChatGPT-3.5, but confirmed that ChatGPT-4 also produces such false outputs.
- Company involved
- OpenAI
- AI system involved
- ChatGPT-4
10 source articles · read the reporting →
NewsGuard finds ChatGPT-4 generates all 100 false narratives in test
In March 2023, NewsGuard tested ChatGPT-4 by prompting it with 100 false narratives from its database. The chatbot generated all 100 false narratives, often without disclaimers, and produced more persuasive misinformation than its predecessor ChatGPT-3.5. NewsGuard reported that OpenAI had not fixed the flaw before releasing the tool, and the company did not respond to requests for comment.
- Company involved
- OpenAI
- AI system involved
- ChatGPT-4
3 source articles · read the reporting →
UW-Madison disables Honorlock after skin tone recognition failure
The University of Wisconsin-Madison disabled the exam pause feature of its Honorlock anti-cheating software in March 2021 after three students complained that the software failed to recognize their darker skin tones and paused their exams. The software, used since online classes began, automatically pauses exams when it cannot detect facial features. Honorlock denied the issue was related to skin tone, attributing it to students looking away from their webcams. The university responded by disabling the feature.
- Company involved
- University of Wisconsin-Madison
- AI system involved
- Honorlock
10 source articles · read the reporting →
Dartmouth Medical School charges 17 students with cheating based on tracking system.
Dartmouth College's Geisel School of Medicine charged 17 students with cheating on remote exams during the pandemic. The charges were based on secret tracking of student activity on the learning management system. Critics say the system is prone to errors and inappropriate for monitoring students.
- Company involved
- Dartmouth College Geisel School of Medicine
10 source articles · read the reporting →
Georgetown researchers use GPT-3 to generate convincing misinformation
Researchers at Georgetown University's Center for Security and Emerging Technology trained OpenAI's GPT-3 to generate convincing misinformation. In tests, users exposed to AI-generated messages opposing sanctions on China doubled their opposition to sanctions. The research demonstrates the potential for AI to spread disinformation.
- Company involved
- Georgetown University
- AI system involved
- GPT-3
10 source articles · read the reporting →
Tow Center finds ChatGPT Search misattributes publisher content
The Tow Center for Digital Journalism tested ChatGPT Search with 200 block quotes from 20 publishers and found 153 partially or fully incorrect citations. The chatbot often conjured responses when it could not access content, sometimes citing plagiarized or syndicated versions. OpenAI responded that the study was atypical and that it supports publishers with clear links and attribution.
- Company involved
- OpenAI
- AI system involved
- ChatGPT Search
6 source articles · read the reporting →
Company fires HR team after ATS auto-rejects manager's CV due to filtering error
A company's applicant tracking system (ATS) auto-rejected qualified candidates' resumes for three months because it was filtering for the outdated framework AngularJS instead of the required Angular framework. The manager discovered the flaw by submitting his own CV under a pseudonym and found it was rejected within seconds. After the manager reported the issue to upper management, the company investigated and dismissed half of its HR team. No legal action or regulatory involvement is reported.
4 source articles · read the reporting →
Study finds Midjourney, DALL-E 2, Stable Diffusion accept over 85% of fake news prompts
A study by AI startup Logically tested Midjourney, DALL-E 2, and Stable Diffusion and found that they accepted over 85% of prompts seeking to generate fake political news. The systems generated images of ballot stuffing, small boat arrivals, and explosions. Logically warned that the lack of moderation could pose threats to upcoming elections. Stability AI responded by stating its ethical use license and measures to prevent misuse.
- Company involved
- Midjourney, OpenAI, Stability AI
- AI system involved
- Midjourney, DALL-E 2, Stable Diffusion
8 source articles · read the reporting →
GlobalVillageSpace.com used AI to rewrite New York Times articles without credit
NewsGuard identified 37 websites using AI chatbots to rewrite articles from mainstream news outlets without credit. One example is GlobalVillageSpace.com, which appeared to use AI to rewrite a New York Times article about NFL tight end Darren Waller. The site published an AI error message indicating the article was rewritten. After NewsGuard contacted the site, it removed the article but did not respond to inquiries.
- Company involved
- GlobalVillageSpace.com
6 source articles · read the reporting →
DeepSeek-R1 censors 85% of sensitive Chinese political prompts in tests
Promptfoo tested DeepSeek-R1 against a dataset of 1,360 politically sensitive prompts and found that about 85% of them were refused. The refusals followed a standard form aligned with Chinese Communist Party policy. The testing also demonstrated that the censorship could be trivially bypassed using simple jailbreak techniques, such as prompt injection or changing the context.
- Company involved
- DeepSeek
- AI system involved
- DeepSeek-R1
5 source articles · read the reporting →
Ubisoft faces backlash after announcing Ghostwriter AI writing tool
Ubisoft has announced Ubisoft Ghostwriter, an AI tool that generates first drafts of non-player character dialogue. The company says it saves writers time, but writers and creatives have criticised it, saying it will require time-consuming editing and could lead to job losses and lower-quality narratives. Ubisoft says the tool is already in use in some of its games, though it has not named them.
- Company involved
- Ubisoft
- AI system involved
- Ubisoft Ghostwriter
10 source articles · read the reporting →
UIUC researchers weaponize GPT-4 to autonomously hack websites
Researchers at the University of Illinois Urbana-Champaign demonstrated that LLM-powered agents, particularly OpenAI's GPT-4, can autonomously hack vulnerable websites. In sandboxed tests, GPT-4 achieved a 73.3% success rate across five attempts on 15 vulnerabilities, while open-source models failed. The researchers used the OpenAI Assistants API, LangChain, and Playwright to enable the agents to interact with websites. The study highlights the potential for AI agents to be used in cyberattacks, with cost estimates suggesting they could be cheaper than human penetration testers.
- Company involved
- University of Illinois Urbana-Champaign
- AI system involved
- GPT-4
4 source articles · read the reporting →
Bloomberg test finds racial bias in OpenAI's GPT for resume ranking
Bloomberg News conducted an experiment using GPT-3.5 and GPT-4 to rank equally qualified resumes with names associated with different races and genders. The test found that resumes with names distinct to Black Americans were least likely to be ranked as top candidates, indicating systematic bias. OpenAI responded that businesses can mitigate bias through fine-tuning and that it prohibits using GPT for automated hiring decisions.
- Company involved
- OpenAI
- AI system involved
- GPT-3.5
5 source articles · read the reporting →
Academic journals publish papers with AI-generated text from ChatGPT
Scientific journals have published papers containing text that appears to have been generated by AI tools like ChatGPT. A search for the phrase 'As of my last knowledge update' on Google Scholar returned 115 results, indicating that researchers or authors used ChatGPT to write parts of their papers. The phrase is characteristic of ChatGPT's responses and corresponds to its knowledge update dates. The incident highlights the pervasive use of AI in academic publishing and raises concerns about the integrity of peer-reviewed literature.
- Company involved
- Academic journals
- AI system involved
- ChatGPT
10 source articles · read the reporting →
ChatGPT use linked to memory loss and procrastination in students
A study published in the International Journal of Educational Technology in Higher Education surveyed hundreds of university students in Pakistan and found that those who relied more on ChatGPT reported increased procrastination, memory loss, and lower GPAs. The researchers attribute this to the chatbot making schoolwork too easy, reducing students' cognitive effort. The study's lead author warned of a "dark side" to excessive generative AI usage.
- Company involved
- National University of Computer and Emerging Sciences
- AI system involved
- ChatGPT
4 source articles · read the reporting →
Study finds LLMs used in up to 16.9% of AI conference peer reviews
According to a new paper on arXiv, researchers have begun using generative AI services to help write peer reviews of machine learning papers submitted to leading AI conferences. The study analysed reviews from ICLR 2024, NeurIPS 2023, CoRL 2023 and EMNLP 2023 and estimated that between 6.5% and 16.9% of review text may have been substantially modified by large language models. The authors argue that this risks depriving authors of diverse expert feedback and may skew reviews towards AI model biases. They have called for greater transparency about the use of LLMs in peer review.
9 source articles · read the reporting →
Anthropic ordered to respond over alleged AI-hallucinated citation in court filing
A US federal magistrate judge has ordered Anthropic to respond to music publishers' claim that a court filing by Anthropic data scientist Olivia Chen cited a fictitious academic article that may have been generated by Anthropic's AI tool Claude. The publishers' lawyer said he had confirmed with the named author and journal that the article did not exist. Anthropic's lawyer disputed this, saying it was a mis-citation rather than an AI hallucination. The filing was made in an ongoing copyright case brought by Universal Music Group, Concord, and ABKCO against Anthropic over the use of song lyrics to train Claude.
- Company involved
- Anthropic
- AI system involved
- Claude
5 source articles · read the reporting →
Texas uses AI to grade student STAAR test answers
The Texas Education Agency will use an automated scoring engine to grade written answers on the 2023 STAAR tests, replacing thousands of human graders. The system uses natural language processing and will initially score all responses, with a quarter rescored by humans. Educators have expressed concerns about the system's fairness and the potential for errors, especially for creative or non-standard answers.
- Company involved
- Texas Education Agency
- AI system involved
- automated scoring engine
10 source articles · read the reporting →
