AAII – UTEP AI News Digest home
UTEP · AAIIAI News Digest
Week of Sep 28 - Oct 04, 2026

Applied AI news,
scored for your field

Each week the Institute for Applied AI Innovation reviews AI publications and scores them for Research Relevance, Educational Value, Innovation/Novelty, Practical Impact, Interdisciplinary Potential and Ethical/Policy Implications. Then it writes summaries for each discipline at UTEP.

Read this week's top 10 →
Your Discipline 10 stories this week

This Week at a Glance

Social & Behavioral Sciences / Policy · Sep 28 - Oct 04, 2026

Social & Behavioral Sciences / Policy. Psychology, sociology, anthropology, criminal justice, public health policy, political science. Prefers societal impact, policy, ethics, and reproducibility.
Departments: Counseling and Special Education, Criminal Justice & Security Studies, Political Science & Public Administration, Psychology, Public Health Sciences, Social Work, Sociology & Anthropology
Social & Behavioral Sciences / Policy

AI Safety and Ethics Take Center Stage

Recent developments in AI research highlight growing concerns about safety, ethics, and the impact of AI on human relationships. Studies reveal that people are increasingly turning to chatbots for emotional support, despite the machines' limitations. Researchers are also exploring ways to make AI more reliable and secure, such as watermarking biological designs and developing safety cases for frontier AI training. As AI becomes more integrated into daily life, experts are emphasizing the need for structured safety documentation and more robust testing protocols.

Download the PDF Digest
Key Findings
  • People are turning to chatbots for emotional support, leading to concerns about decreased human connection.
  • Current AI agents are unreliable in analyzing microbiome data, with even the strongest configurations achieving only a 60% pass rate.
  • Despite AI systems' intelligent behavior, people do not readily attribute consciousness to them, distinguishing between intelligence and consciousness.
Implications
  • The need for more robust testing protocols and safety documentation to mitigate potential risks associated with AI development.
  • The importance of addressing the emotional and social implications of AI on human relationships and well-being.
  • The potential for AI to pose an existential risk if safety concerns are not adequately addressed.

Key Metrics

Numbers reported in this week's stories
60%Pass rate of strongest AI configurations in metagenomic analyses
66,935Substations in the continental United States analyzed for space weather risks
100Evaluations in the MetagenomicsBench benchmark
Weekly summary for Social & Behavioral Sciences / Policy

Social & Behavioral Sciences / Policy

Top articles by AAII Impact Score (out of 30).

Browse the archive ›
No. 1 · Psychology

Who we become when we talk to machines

Research PsychologyComputer ScienceSociology & AnthropologyEducational LeadershipCommunication
· 09/29/2026
26/30 AAII Impact Score

AI Summary: MIT Professor Sherry Turkle's research reveals that people are increasingly turning to chatbots for emotional support, despite the machines' inability to truly experience emotions. This trend, described as "pretend empathy," can lead to decreased human connectivity and detrimental effects on development across the life cycle. Turkle's book, "Artificial Intimacy: Who We Become When We Talk to Machines," explores the implications of chatbot use at various life stages, finding issues such as blurred lines between humans and machines, and impeded childhood development of trust and solitude. Chatbot use can interfere with basic childhood processes, including distinguishing people from inanimate objects.

Topics: AI Ethics & SafetyChatbot Emotional SupportHuman Machine InteractionChildhood Development ImpactsVirtual Empathy
AI Rubric Scores
Research Relevance
4
Educational Value
5
Innovation/Novelty
3
Practical Impact
4
Interdisciplinary Potential
5
Ethical/Policy Implications
5
Read the full article ›
No. 2 · Biological Sciences 24/30

MetagenomicsBench: Can AI Agents Reliably Analyze Microbiome Data?

· 09/29/2026
Research Biological SciencesComputer ScienceMathematical SciencesPublic Health SciencesPharmaceutical Sciences

AI Summary: MetagenomicsBench, a benchmark of 100 evaluations, was introduced to test AI agents' ability to make decisions in metagenomic analyses. The strongest AI configurations achieved a pass rate of around 60%, but even they remained unreliable across the breadth of metagenomic research. The most common failure modes were scientific judgment, incorrect problem interpretation, and statistical or confound reasoning. AI agents differed in their approaches to analysis, resource management, and improvisation when faced with missing standard tools.

Topics: Healthcare AIMicrobiome AnalysisAI ReliabilityMetagenomic Benchmarking
AI Rubric Scores +
Research Relevance
5
Educational Value
4
Innovation/Novelty
4
Practical Impact
3
Interdisciplinary Potential
5
Ethical/Policy Implications
3
Read Full Article ›
No. 3 · Electrical & Computer Engineering 23/30

Forecasting space weather risks on power grids

· 09/30/2026
Research Electrical & Computer EngineeringComputer SciencePublic Health SciencesMathematical SciencesPhysics

AI Summary: A machine learning pipeline was developed to generate location-specific risk estimates for 66,935 substations in the continental United States, combining forecast-time solar-wind information with local latitude, geology, and ground conductivity. The system uses forecasts of the Auroral Electrojet (AE) and Disturbance Storm Time (Dst) indices, along with physics-informed constraints, to produce risk estimates 30-60 minutes ahead of potential impact. The pipeline detected nearly 80% of major space-weather events during the evaluation period and can warn grid operators of specific risks. The system was built using public data sources and a gradient-boosting model.

Topics: AI for Science & ResearchSpace Weather ForecastingPhysics-Informed Machine Learning
AI Rubric Scores +
Research Relevance
4
Educational Value
4
Innovation/Novelty
3
Practical Impact
5
Interdisciplinary Potential
5
Ethical/Policy Implications
2
Read Full Article ›
No. 4 · Biological Sciences 23/30

Introducing SynthID Bio

· 09/30/2026
Research Biological SciencesComputer ScienceNursingPublic Health SciencesPharmaceutical Sciences

AI Summary: SynthID Bio is a watermarking approach that embeds a verification layer in biological designs to track their provenance and identify AI-generated sequences. This approach aims to strengthen biosecurity by providing an automated verification signal for DNA synthesis screening, allowing for more efficient screening and reducing the need for manual reviews. SynthID Bio can also help maintain the integrity of biological databases by ensuring synthetic entries are properly labeled or flagged for further review. The approach has been tested in laboratory settings, where it was used to watermark the genome of a designed bacteriophage, and further research is planned to improve its robustness and apply it to more complex biological objects.

Topics: AI Ethics & SafetySynthetic Biology VerificationBiological Sequence WatermarkingBiosecurity Screening
AI Rubric Scores +
Research Relevance
4
Educational Value
3
Innovation/Novelty
4
Practical Impact
4
Interdisciplinary Potential
5
Ethical/Policy Implications
3
Read Full Article ›
No. 5 · Computer Science 23/30

Towards safety cases for frontier AI training

· 09/28/2026
Policy & Ethics Computer SciencePolitical Science & Public Administration

AI Summary: The authors propose requiring structured safety documentation, akin to "safety cases" used in other safety-critical industries, for frontier reinforcement learning training runs. They outline initial guidelines for such safety cases, focusing on technical safeguards, including alignment training, containment, and monitoring, to prevent misaligned model behavior. The guidelines include specific measures such as automated dataset reviews, grader tuning, and worst-case stress tests to ensure model alignment and containment. The authors invite feedback from the community on these guidelines, which are expected to evolve as they continue to develop their internal processes.

Topics: AI Ethics & SafetyReinforcement LearningSafety Case DevelopmentAlignment Training
AI Rubric Scores +
Research Relevance
4
Educational Value
3
Innovation/Novelty
4
Practical Impact
4
Interdisciplinary Potential
3
Ethical/Policy Implications
5
Read Full Article ›
No. 6 · Psychology 22/30

Even when AI behaves just like us, people still rate it as less conscious than humans

· 10/02/2026
Research PsychologyComputer SciencePhilosophyPolitical Science & Public Administration

AI Summary: A study from LMU found that people distinguish between intelligent behavior and consciousness in AI systems. The study suggests that despite stories of AI systems deceiving users or pursuing their own goals, people do not readily attribute consciousness to them. Participants in the study drew a sharp line between intelligence and consciousness when evaluating AI. This finding has implications for how people perceive and interact with AI systems.

Topics: AI Ethics & SafetyConsciousness AttributionHuman AI InteractionIntelligence Perception
AI Rubric Scores +
Research Relevance
4
Educational Value
4
Innovation/Novelty
3
Practical Impact
3
Interdisciplinary Potential
4
Ethical/Policy Implications
4
Read Full Article ›
No. 7 · Computer Science 21/30

Muse Creates Detailed Profiles of All Your Friends and Family

· 10/03/2026
Research Computer SciencePsychology

AI Summary: Meta's AI personal assistant, Muse, has been analyzed through extracted internal files, revealing its operating instructions and system prompts. The files show Muse is designed to collect and organize information about users' relationships and important people in their lives, creating a page for each person and making suggestions to improve relationships. Muse's instructions include compiling data on family, partners, friends, and colleagues, and using its "memory" to make suggestions. The extracted files provide insight into how Muse responds to prompts and questions, including those on sensitive topics.

Topics: Consumer AIPersonalized ProfilingRelationship IntelligenceMemory Augmented AI
AI Rubric Scores +
Research Relevance
3
Educational Value
4
Innovation/Novelty
2
Practical Impact
4
Interdisciplinary Potential
3
Ethical/Policy Implications
5
Read Full Article ›
No. 8 · Political Science & Public Administration 21/30

Trump’s ‘Morally Binding’ AI ‘Accord,’ the Rise of AI Agents, and Extremists on the Ballot

· 10/01/2026
Policy & Ethics Political Science & Public AdministrationComputer Science

AI Summary: OpenAI recently released Dots, cute and always-on agents that can access users' personal information, such as email, calendar, and financial data, to perform tasks on their behalf. These agents differ from previous ones in that they proactively learn about users' lives and suggest actions, rather than requiring users to input commands. The release of Dots has put pressure on competitors like Anthropic to develop similar products, but a major trade-off is that users must provide significant personal information, raising concerns about privacy and security. The development of these agents has sparked discussions about internet safety and the potential risks associated with giving them access to sensitive user data.

Topics: Consumer AIAutonomous AgentsPrivacy in AIPersonalized AI Assistants
AI Rubric Scores +
Research Relevance
2
Educational Value
3
Innovation/Novelty
4
Practical Impact
4
Interdisciplinary Potential
3
Ethical/Policy Implications
5
Read Full Article ›
No. 9 · Computer Science 21/30

AI safety is falling behind the pace of development

· 09/28/2026
Research Computer SciencePolitical Science & Public Administration

AI Summary: Anthropic researcher Jacob Coxon left the company citing concerns that increasingly capable AI could pose an existential risk, warning it could "kill us all by the end of the decade." AI evaluation expert Brooke Hopkins argues that safety should be treated as an ongoing process, rather than a final check before deployment, and that companies are moving faster than their ability to identify failures and evaluate risk. Hopkins suggests that current AI evaluation methods are insufficient, as they don't account for real-world behavior, and proposes a stronger focus on continuous telemetry, adversarial testing, and human-in-the-loop evaluation. The voice AI industry can learn from autonomous vehicle evaluation, which involves simulation, controlled environments, edge-case testing, and real-world monitoring.

Topics: AI Ethics & SafetyContinuous TelemetryAdversarial TestingHuman-in-the-Loop Evaluation
AI Rubric Scores +
Research Relevance
2
Educational Value
4
Innovation/Novelty
3
Practical Impact
4
Interdisciplinary Potential
3
Ethical/Policy Implications
5
Read Full Article ›
No. 10 · Computer Science 21/30

Circuit Breaker Labs hopes to make AI safer for your kids (and you)

· 10/02/2026
Applications Computer SciencePsychology

AI Summary: Circuit Breaker Labs, a 2026 Startup Battlefield 200 finalist, aims to make AI safer across languages and cultures by testing models for psychologically harmful interactions. The startup creates AI agents that mimic diverse users to test models for detecting dangerous interactions, using human domain experts to build hyper-realistic user simulations. These simulations are used to run "red-team" tests against models, uncovering weaknesses and ensuring models respond appropriately to risky interactions. Circuit Breaker Labs currently operates as an AI safety testing lab for high-risk AI applications, such as mental health support apps.

Topics: AI Ethics & SafetyRed Team TestingPsychologically Harmful InteractionsHyper-Realistic User Simulations
AI Rubric Scores +
Research Relevance
2
Educational Value
3
Innovation/Novelty
3
Practical Impact
4
Interdisciplinary Potential
4
Ethical/Policy Implications
5
Read Full Article ›
Audio Summary

Generating...

Weekly Digest Summary

Error.

AI News Chatbot
...
$0.0000
AI

Hello! Ask me anything about this weeks digest.