What is AIWiki Malaysia?

AIWiki Malaysia is a free, open AI knowledge base covering artificial intelligence concepts, tools, models, and use cases — written specifically for Malaysian professionals and students. It is maintained by AITG Sdn Bhd, an AI company based in Penang.

Who maintains AIWiki Malaysia?

AIWiki Malaysia is maintained by AITG Sdn Bhd (Registration: 202601016521 (1678618-W)), an AI company headquartered in George Town, Penang, Malaysia. The editorial team continuously updates and expands the knowledge base.

What topics does AIWiki Malaysia cover?

AIWiki Malaysia covers a wide range of AI topics including large language models (LLMs), AI agents, machine learning fundamentals, prompt engineering, AI automation, generative AI tools, Malaysian AI regulations, local vendor landscape, and real-world AI use cases relevant to the Malaysian market.

How do I search for AI topics on AIWiki Malaysia?

You can use the search bar at the top of the site to find articles by keyword or topic. Articles are also organised by category, so you can browse by subject area such as Models, Tools, Concepts, or Use Cases.

Is AIWiki Malaysia available in Bahasa Malaysia?

Yes. AIWiki Malaysia publishes content in both English and Bahasa Malaysia to serve the full breadth of the Malaysian professional and student community. Language availability is indicated on each article page.

How can I submit a topic or suggest an article?

You can suggest topics or submit article ideas by contacting the AIWiki Malaysia team at admin@aiteragrid.com. AITG Sdn Bhd reviews all submissions and publishes content that meets editorial accuracy standards.

Anthropic

Anthropic is an American AI safety company and large language model developer founded in 2021 by former OpenAI researchers, best known for developing the Claude family of AI assistants and the Constitutional AI alignment technique.

7 min readLast updated May 2026Companies & Tools

Anthropic is an American artificial intelligence safety company and large language model developer, incorporated as a Public Benefit Corporation (PBC). The company was founded in 2021 by Dario Amodei (Chief Executive Officer), Daniela Amodei (President), and several colleagues who had previously worked at OpenAI, including Tom Brown, Chris Olah, Sam McCandlish, Jack Clark, and Jared Kaplan. Anthropic's stated mission is the responsible development and maintenance of advanced AI for the long-term benefit of humanity, with AI safety research occupying a central role alongside product development.

The company is best known for the Claude family of AI assistants, deployed through a consumer web interface and an API platform. Anthropic has grown rapidly: in February 2026, it announced a USD 30 billion funding round that valued the company at approximately USD 380 billion, with annualised revenue reported to be growing at approximately 10 times year-over-year.[^1]

Founding and Background

Anthropic was established following the departure of a significant cohort from OpenAI in late 2020 and 2021, amid disagreements over the direction and governance of AI development. Dario Amodei had served as VP of Research at OpenAI. The founding team brought deep expertise in large-scale model training, interpretability research, and alignment theory.

The company received early-stage backing from Spark Capital and subsequently from a broader investor group. Amazon Web Services announced a multi-year investment of up to USD 4 billion in Anthropic in 2023, making it the largest external investor and establishing a strategic partnership under which Claude models are available through Amazon Bedrock. In October 2025, Anthropic announced access to up to one million of Google's Tensor Processing Units (TPUs), providing substantial additional compute capacity.[^2]

Claude Models

Claude is Anthropic's primary product and the external manifestation of its research agenda. The model family has evolved through multiple generations.

The first version of Claude was completed in 2022 but not publicly released at that time, with Anthropic citing the need for additional safety evaluation. Claude was launched in March 2023, available to early API customers.

Claude 3 (March 2024) introduced a three-tier capability structure: Opus (highest capability), Sonnet (balanced performance and efficiency), and Haiku (fastest and most compact). Claude 3 Opus demonstrated performance competitive with GPT-4 on standard benchmarks at the time of release.

Claude 4 was announced in May 2025, introducing further capability advances alongside the Model Context Protocol (MCP), an open standard enabling Claude to connect with external tools, databases, and services. By February 2026, the Claude Opus 4.6 and Sonnet 4.6 variants had been released.[^1]

Claude models are deployed through Claude.ai (a web and mobile interface), the Anthropic API, and through third-party cloud platforms including Amazon Bedrock, Google Cloud Vertex AI, and integrations within products such as Slack (Claude for Teams), Notion, and enterprise software ecosystems.

Constitutional AI and Alignment Research

Anthropic's most distinctive technical contribution to the AI safety field is Constitutional AI (CAI), introduced in a 2022 paper.[^3] Constitutional AI is a training methodology designed to produce AI models that are helpful, harmless, and honest without relying entirely on human feedback for every safety-related judgement.

The technique operates in two phases. In the first phase, a model critiques and revises its own outputs according to a set of principles (the "constitution") covering harmlessness, honesty, and avoidance of harmful content. In the second phase, reinforcement learning from AI feedback (RLAIF) is used: a second AI model evaluates competing responses according to constitutional principles and generates preference data used for fine-tuning. This approach reduces dependence on large volumes of human labellers for safety feedback and allows the model's values to be specified and audited through the constitution document rather than inferred from human labelling patterns alone.

Anthropic also conducts foundational research in mechanistic interpretability — the attempt to reverse-engineer the computations performed inside neural networks at the level of individual neurons, circuits, and attention heads. The goal is to eventually make it possible to verify that a model's internal reasoning is aligned with its stated behaviour, rather than relying solely on behavioural evaluation.

Safety Research and Policy

Anthropic publishes an annual Responsible Scaling Policy (RSP) that commits the company to specific safety evaluations before deploying models above defined capability thresholds. The policy establishes AI Safety Level (ASL) tiers: current models are evaluated against criteria for ASL-2 and ASL-3, with higher levels reserved for models that could pose catastrophic risks.

The company maintains a dedicated policy team and engages with legislative processes in the United States, the United Kingdom, and the European Union. Anthropic's researchers have contributed to public debate on frontier AI risk, model evaluations, and standards for responsible deployment.[^4]

Commercial Position

Anthropic competes directly with OpenAI (GPT model family), Google DeepMind (Gemini family), and Meta AI (Llama open-weight models) in the large language model market. Its differentiation rests on its safety-oriented brand positioning, Constitutional AI research, and enterprise adoption among customers for whom AI governance and auditability are procurement priorities. In December 2025, Anthropic announced a USD 200 million multi-year partnership with Snowflake to distribute Claude through Snowflake's data cloud platform.

Malaysian Context — Claude and Anthropic's Presence in Malaysia

Anthropic's Claude models are accessible to Malaysian individuals and organisations through Claude.ai and the Anthropic API, and through Amazon Bedrock — which is available in the AWS Asia Pacific (Singapore) region, the primary AWS region serving Malaysia. Several Malaysian enterprises and technology companies have adopted Claude through Amazon Bedrock for use cases including customer service automation, legal document review, financial report analysis, and internal knowledge management.

Teragrid AI, an AI solutions provider headquartered in Penang, Malaysia, operates the Teragrid Agent platform — an enterprise AI agent system built on the Claude API. The platform deploys Claude's reasoning and instruction-following capabilities for enterprise clients across Malaysia and the broader ASEAN region, integrating Claude with local business processes, Bahasa Malaysia language requirements, and industry-specific regulatory contexts including those governed by Bank Negara Malaysia (BNM) and the Securities Commission Malaysia (SC).

Malaysia's financial services sector has been an early adopter of Claude-based AI tools. Fintech companies operating under BNM's regulatory sandbox, and established financial institutions seeking to deploy AI assistants for internal compliance, audit support, and customer communications, have evaluated and deployed Claude models, attracted by Anthropic's transparency in safety methodology — a priority consideration for institutions regulated under frameworks that require explainability and accountability in automated decision-making.

MDEC and Malaysia's National AI Office, in their engagements with the global AI industry, have noted Anthropic's Constitutional AI methodology as an example of responsible AI development practice aligned with the principles of Malaysia's AI Governance Framework. Anthropic's approach of publishing formal safety commitments through its Responsible Scaling Policy provides a template that Malaysian regulators have referenced in discussions about how AI developers should document and communicate safety practices.

The model context protocol (MCP) introduced by Anthropic in 2025 has been adopted by Malaysian AI developers as a standard for connecting Claude to enterprise data sources, enabling Malaysian-built tools and connectors to integrate with Claude in a standardised way — supporting the growth of a local AI solutions ecosystem around the Claude platform.

As AI skills development accelerates in Malaysia through HRD Corp-approved programmes and university curricula, Anthropic's published research on alignment, interpretability, and safety evaluation is increasingly referenced as required reading for advanced AI practitioners, alongside the work of OpenAI, DeepMind, and academic institutions.

References

Taskade. (2026). Anthropic History 2026: Claude AI to $380B Valuation. https://www.taskade.com/blog/anthropic-claude-history
Britannica Money. (2025). Anthropic | History, Controversies, & Claude AI. Encyclopaedia Britannica. https://www.britannica.com/money/Anthropic-PBC
Bai, Y., Jones, A., Ndousse, K., et al. (2022). Constitutional AI: Harmlessness from AI Feedback. arXiv:2212.08073. Anthropic.
Anthropic. (2023). Anthropic's Responsible Scaling Policy. https://www.anthropic.com/responsible-scaling-policy

Tags:AI safety Claude large language models Constitutional AI alignment

Type	Private AI safety company (PBC)
Founded	2021
Headquarters	San Francisco, California, USA
CEO	Dario Amodei
Key product	Claude (AI assistant and API)
Valuation	Approximately USD 380 billion (February 2026)

Founding and Background

Claude Models

Constitutional AI and Alignment Research

Safety Research and Policy

Commercial Position

See Also

References

References