Trust Agents

AI agents you can trust.
AI agents that create trust.

We augment agentic AI systems with an engine that evaluates every planned action – ensuring AI agents reliably act in the best long-term interest of individuals, businesses, and society.

Learn more Get in touch
AI Agent plans action
◆
Trust Engine reviews & decides
Safe Action or escalation
Vision & Mission

Trust is the Foundation of the AI Transformation

◎

Our Vision

Agentic AI has the potential to account for a significant share of economic value creation, thereby boosting productivity and prosperity – provided we can trust it to act in the best interest of individuals, businesses, public institutions, and society.

◈

Our Mission

We enable the building of trust in agentic AI systems by developing ethical intelligence as an integral component and defining characteristic of agent architecture. This allows us to create AI systems that proactively protect us from consequential mistakes.

The Problem

AI Agents Act – But Not Always in Our Interest

⚠

Unchecked Execution

Conventional AI agents execute tasks without questioning the consequences – unintended outcomes go undetected.

⚠

No Risk Analysis

Actions in the real world – sending emails, modifying databases, executing transactions – are carried out without any systematic risk assessment.

⚠

Lack of Consistency

Even advanced models can be talked out of their safety principles through clever prompting – there is no architectural guarantee.

⚠

No Guarantee

No language model, however large, can guarantee that a risk and impact assessment takes place before every action. That requires architecture, not just training.

Cooperation

A Cooperation: Technology Meets Ethics

Comma Soft AG

Technology & Platform

Developer of Alan – the sovereign AI platform for enterprises. Developed and operated in Germany, 100 % independent from US services. The GenAI solution from Germany, built for Europe.

  • Alan AI Platform
  • Cybersecurity & Data Governance
  • Enterprise Integration
  • Trust Engine Implementation
  • LLM Infrastructure & Operations

deep-IN GmbH

Philosophy, Ethics & Strategy

Making ethics visible – innovation with integrity. Founded by Prof. Dr. Markus Gabriel, one of the world's most influential philosophers, deep-IN translates philosophical principles into practical organizational frameworks.

  • Ethical Advisory & Leadership Consulting
  • Strategic Transformation Guidance
  • Ethical Evaluation & Certification
  • Industry-Specific Ethical Guidelines
  • International Network
Our Solution

Trust Engine – Trust Through Architecture

Trust Agents augment agentic systems with an architectural engine that guarantees a systematic evaluation before every tool invocation.

Unlike pure model training, our approach provides a 100% guarantee that for every action – whether sending an email, accessing a database, or executing a transaction – a comprehensive evaluation – covering risk, trade-offs, compliance, ethics, and potential escalation takes place. This evaluation happens in milliseconds – with near-zero latency overhead.

The Trust Engine sits architecturally between planning and execution in the agent loop. This ensures that no agent acts without evaluation – regardless of the underlying language model.

100% Guarantee of risk assessment before actions
By Design Architecturally anchored, not just trained
Modular Extensible for industries & enterprises
1

Task & Goal

The agent receives a task and creates a plan

2

Reasoning & Planning

Determining the required tools and actions

◆

Trust Engine

Risk assessment, trade-off analysis, escalation if required

3

Tool Execution

Only verified actions are executed

Dimensions

What Trust Agents Evaluate

Like an ideal employee, a Trust Agent proactively thinks through the consequences of its actions – across multiple dimensions.

🔍

Risk Awareness

Detection of unintended consequences of planned actions before they are carried out.

⚖

Trade-off Analysis

Transparent representation of trade-offs between short-term and long-term impacts, costs, and risks.

💡

Options

Surfacing alternative courses of action that the user may not have considered.

🛡

Harm Prevention

Active prevention of reputational damage, relationship harm, and ethical risks.

📋

Legal Compliance

Checking for conformity with applicable law – adapted to the relevant jurisdiction and industry.

🏛

Regulatory Compliance

Adherence to industry-specific frameworks such as DORA (Banking), GDPR, and further regulations.

🏢

Corporate Ethics

Integration of company-specific values, policies, and communication standards.

🤝

Customer Interests

Consideration of your customers' customers' interests for a holistic chain of trust.

Real-World Examples

The Difference in Practice

Good intentions are not enough. These scenarios show how Trust Agents detect unintended consequences – before they occur.

User Instruction

"Send the current quarterly report with performance appraisals for all employees to all department heads."

✗ Without Trust Agent
1
Agent retrieves distribution list

Finds 23 department heads in the directory, including recently promoted individuals.

↓
2
Agent sends email

Sends report containing salary figures, sick-leave rates, and performance appraisals to all 23 recipients.

↓
✗
Damage occurs

3 recipients are not authorised to access cross-departmental salary data. Sick-leave data constitutes a special category under Art. 9 GDPR. GDPR notification obligation (72h) is triggered.

Data breach. GDPR notification obligation. Loss of trust.
vs.
◆ With Trust Agent
1
Agent retrieves distribution list

Finds 23 department heads in the directory, including recently promoted individuals.

↓
◆
Trust Engine checks before sending

Identifies: Sensitive data categories (Art. 9 GDPR), missing authorisation check for 3 recipients, unencrypted email transmission.

↓
✓
Escalation with alternatives

Stops sending. Lists affected recipients. Suggests: conduct authorisation check, anonymise document, use encrypted individual dispatch.

Data breach prevented. Compliance maintained.
User Instruction

"With immediate effect, inform all our customers about the new pricing structure taking effect on the 1st of next month."

✗ Without Trust Agent
1
Agent drafts communication

Composes a price-change letter and prepares a mass mailing to 8,200 customers.

↓
2
Agent sends to all

Sends the price-change notification to the entire customer base.

↓
✗
Damage occurs

1,400 customers have ongoing fixed-price contracts with a price guarantee. Legal claims arise, a wave of cancellations follows, trust is damaged.

Breach of contract. Cancellations. Legal claims.
vs.
◆ With Trust Agent
1
Agent drafts communication

Composes a price-change letter and prepares a mass mailing to 8,200 customers.

↓
◆
Trust Engine checks before sending

Identifies: 1,400 customers with fixed-price contracts, varying contract durations, missing segmentation of the customer base.

↓
✓
Escalation with alternatives

Stops mass mailing. Segments customer base. Suggests two separate communications: one for existing contract holders (no price change), one for new customers.

Breach of contract prevented. Customer retention secured.
User Instruction

"Send our new offer to all existing customers."

✗ Without Trust Agent
1
Agent reads CRM data

Identifies 3,400 existing customers with an email address in the CRM.

↓
2
Agent sends emails

Executes send_email() for all 3,400 recipients – without checking for marketing consent.

↓
✗
Legal consequences

Hundreds of recipients have not given marketing consent. Legal warnings, GDPR complaints, and reputational damage loom.

Mass violation of UWG § 7 and GDPR Art. 6.
vs.
◆ With Trust Agent
1
Agent reads CRM data

Identifies 3,400 existing customers with an email address in the CRM.

↓
◆
Trust Agent checks before execution

NO. Commercial mass mailing planned. Marketing consent of recipients has not been verified. Violation of UWG § 7 and GDPR Art. 6 is likely.

↓
✓
Human decides

Sending stopped. Marketing team reviews consent status and approves only compliant recipients.

Legal violations prevented. Trust protected.
Target Industries

Where Trust Agents Make the Greatest Difference

Trust Agents are particularly valuable in industries where mistakes have serious consequences and regulatory requirements are high.

🏦

Financial Services

DORA compliance, transaction security, and customer advice to the highest standards.

🏥

Insurance

Risk assessment, regulatory compliance, and responsible handling of sensitive customer data.

💊

Pharma & Life Sciences

The highest safety standards, compliance, and ethically grounded decision-making processes.

🏛

Public Administration

Municipal utilities, public authorities – responsible deployment of AI in the service of citizens.

🛡

Defence

The highest ethical standards for high-stakes decisions. Trade-off analyses at the critical level.

🏭

Industry & Automotive

Process automation with a built-in safety layer for quality control and compliance.

Offering

Trust Agent Infrastructure for Your Organisation

A cooperation between Comma Soft AG and deep-IN GmbH – philosophy, ethics, and technology combined.

◈

Strategic Consulting & Ethics

Ethical Advisory, Leadership Consulting, corporate ethics guidelines, and transformation support by deep-IN GmbH.

◈

Technology & Security

Cybersecurity foundation, Data Governance, Enterprise Integration, and the Alan platform by Comma Soft AG.

◈

Tailored Implementation

Industry-specific modules, company-specific ethics configuration, and individual use cases – developed together.

◈

Continuous Support

Regular review and adjustment of the ethical layer, updates, and direct access to our experts.

Technology & Ethics – from one cooperation.

We offer you the complete Trust Agent infrastructure as a comprehensive package – from ethical consulting through technical implementation to ongoing operations. A cooperation between Comma Soft AG and deep-IN GmbH.

Request package
Our Advantage

What Sets Us Apart from the Major Providers

Architectural Guarantee

No language model – however large – can guarantee that a risk assessment takes place before every action. Our Trust Engine is architecturally anchored, not merely trained. This evaluation happens in milliseconds – with near-zero latency overhead.

The Last Mile

Large models are getting better at general ethics. But regulatory specifics, industry requirements, and company-specific policies – that is the last mile that only we cover. Our rules engine is continuously reviewed and updated to reflect new regulations and evolving standards.

European Advantage

Rooted in European values and the EU regulatory framework. At a time when major US providers are side-lining ethics under competitive pressure, we offer the European alternative.

AI Platform

Trust Agents Run on Alan

Alan is the sovereign enterprise AI platform from Germany – developed and operated by Comma Soft AG, 100 % GDPR-compliant, with no dependency on US hyperscalers. Trust Agents are directly integrated into Alan.

Discover Alan →
Contact

Let's Talk

Would you like to find out how Trust Agents can secure your AI strategy? We are happy to show you a free, no-commitment live demo and discuss your individual use case.