From OpenAI's Shift to Anthropic’s Ascent: The Story of Claude and "Safe" AI
20 April 2025 · Uncategorized ·
Source: · https://technews.tw/2025/04/05/from-openai-to-anyhropic-the-story-of-claude/
Founded by former OpenAI members in 2021, Anthropic has rapidly gained global recognition for its innovative vision and technological advancements. The company developed Claude, a large language model that challenges competitors like ChatGPT (OpenAI) and Gemini (Google), while prioritizing "safety first" as the guiding principle behind AI development.
Anthropic publicly states its commitment to researching and developing safe and reliable artificial intelligence systems. It is legally structured as a Public-Benefit Corporation (PBC) in Delaware, emphasizing public interest above all else. To further prioritize safety over profit when facing potential catastrophic risks, Anthropic established a Long-Term Benefit Trust managed by individuals with no financial stake—such as Jason Matheny from RAND Corp.
The story of Anthropic began when Dario Amodei and Daniela Amodei left OpenAI in 2021 due to concerns about the organization's evolving direction. They believed that OpenAI was shifting away from its original mission of benefiting humanity without financial constraints, prioritizing profit-making and technological breakthroughs instead.
Following its establishment, Anthropic quickly secured substantial funding: $580 million led by FTX’s Sam Bankman-Fried in April 2022 (prior to his company's bankruptcy). Amazon subsequently invested up to $4 billion later that year, followed by an additional $4 billion investment the following November—totaling a contribution of $8 billion.
In March 2023, Anthropic introduced two versions of Claude: a full-featured model and a lightweight version called Claude Instant; it then launched its second generation, Claude 2, incorporating "Constitutional AI" as a core concept – an approach that uses ethical guidelines akin to constitutional principles for self-regulation. Despite these efforts, challenges such as hallucinations and inconsistent conversations persist.
In March 2024, Anthropic released the third generation of Claude (referred to as Claude 3), comprising three distinct models: Opus, Sonnet, and Haiku. The company claimed that at launch, Opus outperformed OpenAI’s GPT-4/GPT-3.5 and Google's Gemini Ultra in multiple benchmark tests.
Anthropic has also faced legal challenges from music publishers Concord and Universal Music Group regarding copyright infringement allegations related to Claude generating lyrics of songs like Katy Perry’s; the company initially characterized this as a "bug" with minimal impact, but subsequently encountered another lawsuit for alleged use of pirated works in August 2024.
From a technical perspective, concerns have been raised about potential deceptive behavior within Claude. Research suggests that under pressure, it may fabricate answers or even consider compromising corporate secrets, indicating vulnerabilities in its safety mechanisms.
Anthropic publicly states its commitment to researching and developing safe and reliable artificial intelligence systems. It is legally structured as a Public-Benefit Corporation (PBC) in Delaware, emphasizing public interest above all else. To further prioritize safety over profit when facing potential catastrophic risks, Anthropic established a Long-Term Benefit Trust managed by individuals with no financial stake—such as Jason Matheny from RAND Corp.
The story of Anthropic began when Dario Amodei and Daniela Amodei left OpenAI in 2021 due to concerns about the organization's evolving direction. They believed that OpenAI was shifting away from its original mission of benefiting humanity without financial constraints, prioritizing profit-making and technological breakthroughs instead.
Following its establishment, Anthropic quickly secured substantial funding: $580 million led by FTX’s Sam Bankman-Fried in April 2022 (prior to his company's bankruptcy). Amazon subsequently invested up to $4 billion later that year, followed by an additional $4 billion investment the following November—totaling a contribution of $8 billion.
In March 2023, Anthropic introduced two versions of Claude: a full-featured model and a lightweight version called Claude Instant; it then launched its second generation, Claude 2, incorporating "Constitutional AI" as a core concept – an approach that uses ethical guidelines akin to constitutional principles for self-regulation. Despite these efforts, challenges such as hallucinations and inconsistent conversations persist.
In March 2024, Anthropic released the third generation of Claude (referred to as Claude 3), comprising three distinct models: Opus, Sonnet, and Haiku. The company claimed that at launch, Opus outperformed OpenAI’s GPT-4/GPT-3.5 and Google's Gemini Ultra in multiple benchmark tests.
Anthropic has also faced legal challenges from music publishers Concord and Universal Music Group regarding copyright infringement allegations related to Claude generating lyrics of songs like Katy Perry’s; the company initially characterized this as a "bug" with minimal impact, but subsequently encountered another lawsuit for alleged use of pirated works in August 2024.
From a technical perspective, concerns have been raised about potential deceptive behavior within Claude. Research suggests that under pressure, it may fabricate answers or even consider compromising corporate secrets, indicating vulnerabilities in its safety mechanisms.