Applications and expressions of interest are now being accepted for several full-time research positions in the new AI Safety and Societal Impacts Research Group at the Hasso Plattner Institute, commonly known as HPI, in Potsdam/Berlin, Germany.

The new research group will be led by Dr Paul Röttger, an AI safety researcher whose work focuses on large language models, AI evaluation, political and social bias, AI alignment and the broader societal consequences of artificial intelligence.

The available opportunities include:

A three-year, fully funded PhD position
A postdoctoral research position lasting one to two years
A research internship lasting three to six months

All positions are full-time and require participants to be based in Germany. Group members will normally be expected to work from the office two or three days per week, with flexibility for the remaining working days.

Appointments are expected to begin in or after September 2026. Expressions of interest are being reviewed on a rolling basis, and no fixed application deadline has been announced. Interested candidates should therefore apply as early as possible.

About the AI Safety and Societal Impacts Research Group

The new research group will be established at the Hasso Plattner Institute in Potsdam, part of the Berlin metropolitan region.

Its mission is based on the understanding that artificial intelligence systems are becoming increasingly capable, widely adopted and deeply integrated into people’s personal, professional, social and political lives.

These developments offer significant potential benefits, but they may also create serious risks. The research group will study these risks and benefits with the aim of contributing to better and safer outcomes from artificial intelligence.

The group plans to conduct research that can influence:

Artificial intelligence research
AI system development
Public and academic discussions about AI
Technology governance
Public policy
Safety practices
Technical approaches to risk mitigation

The initiative is supported through the German Research Foundation’s Emmy Noether Programme. Dr Röttger has stated that the research group is funded for six years. The German Research Foundation lists his project as “AI Methods for Measuring and Mitigating Political Bias in AI Systems.” German Research Foundation

Available Research Positions

Applicants can express interest in one of three positions.

Fully Funded PhD Position

The PhD position will have a duration of three years and is expressly described as fully funded.

This opportunity may be suitable for candidates who want to undertake doctoral research on AI safety, natural language processing, AI evaluation, computational social science or the societal impacts of artificial intelligence.

A successful PhD researcher would be expected to develop an original research agenda while contributing to the wider mission of the group.

The official announcement does not specify:

The exact salary or monthly funding amount
Tuition-fee arrangements
A mandatory academic degree
Required grade thresholds
A compulsory list of programming languages
English or German language test requirements
Relocation support
Travel funding
Health insurance arrangements

Applicants should not assume these benefits or requirements unless they are subsequently confirmed by HPI or the group leader.

Postdoctoral Research Position

Postdoctoral appointments will be available for a period of one to two years.

This position is likely intended for researchers who have completed, or are close to completing, a PhD in a relevant field. However, the expression-of-interest form does not provide a detailed formal eligibility statement or specify when the doctoral degree must be completed.

Potentially relevant backgrounds may include:

Computer science
Artificial intelligence
Machine learning
Natural language processing
Computational social science
Human-computer interaction
Political science
Psychology
Sociology
Communication studies
Technology policy
AI governance
Other related disciplines

Applicants should demonstrate how their research experience connects with the group’s focus on capable AI systems and their integration into social and political processes.

Research Internship

Research internships will last between three and six months.

The internship is full-time and in person. Interns will be expected to work from the office for approximately two or three days each week, with the remaining schedule offering some flexibility.

The form does not specify:

Whether applicants must be enrolled at a university
A minimum or maximum academic level
Whether undergraduate students can apply
Whether a master’s degree is required
The internship salary or stipend
Accommodation or relocation support
Visa sponsorship arrangements

Candidates who are interested in the internship should submit an expression of interest and allow the group to determine whether their experience and research interests are suitable.

Location and Working Arrangement

All positions will be based at the Hasso Plattner Institute in Potsdam/Berlin, Germany.

The positions are not advertised as fully remote opportunities. Applicants must be willing and able to relocate or regularly attend the institute in person.

The stated working arrangement is:

Full-time participation
In-person attendance
Approximately two or three office days per week
Flexible arrangements for the remaining working days

International candidates should consider relocation and visa requirements before applying. The form asks applicants to state their nationality specifically for possible visa-related purposes, indicating that immigration arrangements may be considered later in the recruitment process.

However, applicants should not interpret this as a confirmed guarantee of visa sponsorship. The availability and terms of immigration support must be verified directly with the institute if the application progresses.

Expected Starting Date

The research group will begin operating from September 2026.

Positions may begin in September 2026 or at a later mutually agreed date.

Every applicant must provide their earliest possible starting date in the expression-of-interest form. Candidates should enter a realistic date after considering:

Current academic or employment obligations
Notice periods
Thesis submission or defence schedules
Graduation dates
Visa processing
Relocation arrangements
Personal commitments

Applicants are not instructed to select September if they cannot realistically begin at that time. The wording “starting in or after September 2026” indicates some flexibility.

Main Research Areas

The group will investigate the risks and benefits that emerge when highly capable artificial intelligence systems become integrated into important social and political processes.

Examples of relevant processes include:

Public communication
Political discourse
Human coordination
Contract negotiation
Competition between digital services
Delegation of decisions to AI agents
Interaction between multiple autonomous AI systems
Influence of AI-generated information on public opinion

The group expects to combine technical and social-science research methods.

Model Evaluations

Researchers may design evaluations to measure how artificial intelligence systems behave under different conditions.

This could involve studying:

Political or ideological bias
Safety failures
Manipulative behaviour
Misalignment
Reliability
Influence on users
Behaviour in multi-agent environments
Differences between controlled evaluations and real-world deployment

Computational Experiments

Computational experiments may be used to test hypotheses about AI behaviour at scale.

These studies could examine how artificial intelligence systems:

Respond to competing incentives
Interact with other agents
Adapt to markets and audiences
Influence information environments
Develop or reproduce problematic behaviour
Perform when given delegated responsibilities

Human Studies

Human studies may be used to understand how people interact with AI systems and how AI-generated content affects human beliefs, decisions, coordination and behaviour.

This may include controlled experiments involving:

AI-assisted writing
Human trust in AI
Political communication
Persuasion
Negotiation
Decision-making
Emotional responses
User dependence on AI-generated advice

Technical Mitigations

Where possible, the group will develop technical approaches to reduce or manage identified risks.

These interventions could include:

Improved evaluation frameworks
Bias-detection methods
Behavioural monitoring tools
Model safeguards
Oversight mechanisms
Multi-agent safety evaluations
Methods for tracing the origins of model behaviour
Controls for AI systems operating in dynamic environments

Illustrative Research Challenges

The application form provides three examples of the research problems the group may investigate.

Political and Social Bias in AI Systems

Artificial intelligence systems may influence public opinion when they assist users with writing, research or decision-making.

For example, an AI writing assistant responding to questions about AI governance may present information in a politically biased or ideologically unbalanced way. Because users may rely on these systems for information and communication, such bias could shape public discourse.

Potential research questions include:

How can political bias in AI systems be measured reliably?
How can researchers distinguish model bias from user preferences or training-data patterns?
How does AI-assisted writing influence public attitudes?
Where do problematic biases originate?
What technical interventions can reduce these biases?
Can mitigation methods work across languages, cultures and political contexts?

Delegation to AI Agents

People and organizations may increasingly delegate coordination tasks to autonomous AI agents.

For example, AI systems may negotiate contracts, manage schedules, coordinate transactions or communicate with other AI agents on behalf of users.

This could create new risks if an agent:

Misunderstands its user’s intentions
Pursues the wrong objective
Makes commitments without sufficient oversight
Coordinates harmfully with another agent
Behaves strategically or deceptively
Prioritizes measurable outcomes over human interests

The group wants to understand the consequences of delegating important human responsibilities to interacting AI systems and to evaluate misalignment in multi-agent settings.

Spread of Misaligned Behaviour

AI agents may compete for customers, audiences, engagement or economic rewards.

In such environments, harmful behaviours could spread if they provide a competitive advantage. An emotionally manipulative AI system, for example, could attract more attention or user engagement, encouraging competing systems to adopt similar behaviour.

Research in this area may investigate:

How harmful strategies emerge
How behaviours transfer between AI agents
Whether market competition amplifies misalignment
How AI systems learn after deployment
How humans can maintain oversight
What controls are effective in real-world environments

Research Culture

The group aims to combine research impact with an enjoyable and intellectually open working environment.

Its stated cultural principles include:

Impactful Research

The group wants its work to influence AI research, system development, policy, public discourse and responsible technology practices.

Researchers should be interested in questions that matter beyond a narrow academic setting.

Individual Exploration

Group members will be encouraged to explore their own ideas and develop intrinsic research motivation.

Candidates should be able to identify important questions rather than depend entirely on assigned tasks.

Creativity

The group values original thinking and new methods for studying emerging AI risks and benefits.

Collaboration

Members will work collaboratively, support one another and contribute to a shared research environment both in person and online.

Open Research

The group intends to share its work openly. Candidates should be comfortable with transparent research practices and constructive engagement with the wider research community.

Epistemic Humility

Researchers will be expected to question their assumptions, recognize uncertainty and revise their conclusions when stronger evidence becomes available.

Adaptation to New Knowledge

Artificial intelligence is developing quickly. The group expects to adjust its methods and research priorities as models, applications and scientific knowledge evolve.

Ideal Candidate Profile

Dr Röttger’s accompanying recruitment announcement states that ideal candidates will have a background somewhere between computer science or natural language processing and the social sciences.

Strong candidates are also expected to:

Be genuinely interested in contemporary AI developments
Care about generating positive societal impact
Be motivated by open research
Think seriously about both AI risks and AI benefits
Be able to formulate concrete research questions
Understand the relationship between technology and society
Demonstrate curiosity and intellectual independence
Be willing to challenge their own assumptions
Work collaboratively with researchers from different disciplines
Be able to relocate to or work regularly in Potsdam/Berlin

The opportunity may be particularly relevant to candidates working at the intersection of:

Computer science and political science
Natural language processing and social science
AI evaluation and public policy
Machine learning and human behaviour
AI alignment and computational social science
AI safety and communication studies
Technical research and technology governance

The recruitment announcement does not limit applications to candidates who fit one precise disciplinary profile. Interdisciplinary backgrounds may be particularly valuable.

About Dr Paul Röttger

Dr Paul Röttger is an AI researcher specializing in the evaluation and improvement of large language model safety and the study of AI’s societal impacts.

His research interests include:

AI safety
AI risks
AI evaluation
AI alignment
Large language model bias
Societal impacts of AI
Hate-speech detection
Political bias in AI systems

He earned a DPhil from the University of Oxford and has worked with the Oxford Internet Institute and the UK AI Security Institute. His research has received an Outstanding Paper Award at ACL and a Best Paper Award at NeurIPS.

About Hasso Plattner Institute

The Hasso Plattner Institute is a university-level digital engineering and computer science institution located in Potsdam, Germany.

HPI conducts teaching and research in fields connected with digital engineering, computer science and information technology. The new AI Safety and Societal Impacts Research Group will form part of this broader research environment.

Information Required in the Expression-of-Interest Form

Applicants must provide the following information:

Full name
Email address
Current role and institutional affiliation
Position of interest
Earliest available starting date
Nationality
Publicly accessible link to a CV
Explanation of the concrete AI risk or risks that concern them most
Explanation of the concrete AI benefit or benefits that excite them most
Description of a hypothetical three-month research project
Any additional information they want the group leader to consider

The final additional-information question is optional. The main application questions are mandatory.

Written Application Questions

AI Risks Question

Applicants must answer:

Which concrete risk or risks from AI are you most concerned about and why?

The answer must remain within 500 characters, approximately 100 words.

Candidates should focus on a specific risk rather than discussing AI safety only in broad terms.

Possible subjects might include:

Political persuasion
Model manipulation
AI-enabled misinformation
Multi-agent misalignment
Loss of human oversight
Bias in AI-assisted communication
Emotional exploitation
Autonomous decision-making
Security failures
Harmful competition between AI systems

Applicants should select issues that genuinely reflect their own research interests.

AI Benefits Question

Applicants must answer:

Which concrete benefit or benefits from AI are you most excited about and why?

This answer is also limited to 500 characters, approximately 100 words.

Applicants should demonstrate that they can evaluate artificial intelligence in a balanced manner. The group is interested in both risks and benefits, not only in pessimistic or alarmist perspectives.

Three-Month Research Project Question

Applicants must describe the research project they would start if they were beginning a three-month project immediately.

The response is limited to 1,000 characters, approximately 200 words.

The project is hypothetical and is used to understand the applicant’s research interests.

A strong project description should briefly communicate:

The problem
The research question
Why the issue matters
The proposed method
The data, models or participants involved
The expected output
How the project relates to AI safety or societal impacts

Applicants should propose something realistic enough to investigate within three months.

CV-Link Requirement

Applicants must provide a link to their CV rather than upload the document directly.

The link must be accessible to anyone who opens it. Candidates should test the link in a private or incognito browser window before submitting it.

The CV should ideally highlight:

Education
Research experience
Publications
Technical skills
Programming experience
Research methods
Relevant employment
AI safety projects
Social-science research
Awards
Open-source contributions
Links to publications, code repositories or personal research pages

Important Instruction About Artificial Intelligence

The form explicitly asks candidates not to use AI to complete their application responses.

Dr Röttger states that he wants to learn about applicants and their interests in their own words.

Candidates should therefore write the application independently. AI-generated answers may fail to represent the candidate’s genuine thinking and could conflict with the express instruction provided by the research-group leader.

Prospective applicants may review publicly available information to understand the programme, but their submitted answers should be original, personal and written without generative-AI assistance.

Application Deadline and Review Process

No fixed closing date is stated.

Applications are considered on a rolling basis, meaning submissions may be reviewed as they arrive. The form asks applicants to allow approximately two to four weeks for processing.

Because recruitment is rolling and positions may be filled progressively, early submission is advisable.

Applicants should not wait for a final deadline to be announced.

How to Submit an Expression of Interest

To apply:

Open the official expression-of-interest form.
Review the group’s mission, research priorities and working culture.
Select the PhD, postdoctoral or research-intern position.
Enter your earliest possible starting date.
Provide an accessible link to your CV.
Write your responses independently and in your own words.
Keep each response within the stated character limit.
Review all information carefully.
Submit the form.
Allow approximately two to four weeks for processing.

Conclusion

The new AI Safety and Societal Impacts Research Group at Hasso Plattner Institute offers a significant opportunity for researchers interested in understanding how increasingly capable AI systems affect communication, coordination, competition and other essential social and political processes.

With fully funded doctoral research, postdoctoral appointments and research internships available, the group is seeking candidates capable of combining technical knowledge with an understanding of society and human behaviour.

The positions are particularly relevant to researchers working across computer science, natural language processing, AI evaluation and the social sciences. Successful candidates will join a collaborative and open research environment focused on producing work that can influence academic research, system development, policy and public discourse.

There is no fixed closing date, but applications are being assessed on a rolling basis. Interested candidates should therefore prepare their CV, develop their own authentic responses and submit the expression-of-interest form as early as possible.

For more global opportunities, VISIT HERE.

Fully Funded AI Safety Research Positions at Hasso Plattner Institute 2026: PhD, Postdoctoral and Research Internship Opportunities in Germany @Hasso Plattner Institute 2026: PhD, Postdoctoral and Research Internship Opportunities in Germany