HomeAnthropic Implements New AI Safeguards Amid Rising Risks
Anthropic has updated the usage policy for its Claude AI chatbot to address growing safety concerns. Alongside stricter cybersecurity rules, the company now explicitly outlines categories of dangerous weapons that must not be developed using Claude.
While the official announcement post does not explicitly highlight the changes to its weapons policy, a comparison between the old and new usage guidelines reveals a significant update. Previously, Anthropic broadly banned using Claude to "produce, modify, design, market, or distribute weapons, explosives, dangerous materials or other systems designed to cause harm to or loss of human life." The revised policy expands this by specifically prohibiting the development of high-yield explosives, as well as biological, nuclear, chemical, and radiological (CBRN) weapons.
In May, Anthropic introduced "AI Safety Level 3" protections concurrent with the launch of its new Claude Opus 4 model. These safeguards are intended to make the model more resistant to jailbreaking and to help prevent it from aiding in the creation of CBRN weapons.
In its announcement, Anthropic also recognizes the risks associated with agentic AI tools like Computer Use—which allows Claude to control a user's computer—and Claude Code, which integrates Claude directly into a developer's terminal. "These powerful capabilities introduce new risks, including potential for scaled abuse, malware creation, and cyber attacks," the company states.
Related
- Anthropic launches new Claude service for military and intelligence use
- Google scraps promise not to develop AI weapons
The AI startup is addressing these potential threats by adding a new section titled "Do Not Compromise Computer or Network Systems" to its usage policy. This section includes rules against using Claude to discover or exploit vulnerabilities, create or distribute malware, develop tools for denial-of-service attacks, and similar activities.
Additionally, Anthropic is easing its policy regarding political content. Instead of banning all content creation related to political campaigns and lobbying, the company will now only prohibit using Claude for "use cases that are deceptive or disruptive to democratic processes, or involve voter and campaign targeting." Anthropic also clarified that its requirements for all "high-risk" use cases—which apply when Claude is used to make recommendations to individuals or customers—are relevant only for consumer-facing scenarios, not for internal business use.
Related article
Barry Diller: Trust in Sam Altman irrelevant as AGI nears
Barry Diller, the billionaire media titan, does not believe OpenAI CEO Sam Altman is untrustworthy, despite recent reports suggesting otherwise. Speaking at the Wall Street Journal's "Future of Everything" conference this week, Diller defended Altman
YouTube expands AI deepfake detection to politicians, government officials, and journalists
On Tuesday, YouTube announced it is expanding its deepfake detection technology to a select group of government officials, political candidates, and journalists. The tool identifies AI-generated likenesses and lets pilot participants request the remo
The Real Difference: Not One Thing, but Another
Sometimes, things are not only one thing but also another. The phrase "It's not just this — it's that" has become so common in AI-generated writing that it now serves as more than a hint of synthetic content — it's nearly a certainty.That's why, when
Related Special Topic Recommendations
writing
Best AI Xianxia & Wuxia Assistants: Write Epic Cultivation Progression & Martial Arts Choreography
Discover the 2026 best AI assistants for crafting epic xianxia & wuxia tales. XIX.AI's curated list features top-rated, game-changing tools to master cultivation progression and martial arts choreography. Compare free vs paid options with real-world tests. Unlock your creative potential and start writing today!
10 tools
xix.ai
code
AI Mobile App Coding Tools: Generate Cross-Platform Flutter & React Native Code from Prompts
Discover the 2026 best AI mobile app coding tools for Flutter & React Native. Our curated, top-rated list features powerful, game-changing solutions that generate cross-platform code from prompts. Compare free vs paid options with real-world tests. Unlock faster development and build better apps. Explore the rankings on XIX.AI now!
10 tools
xix.ai
code
Best AI Chrome Extension Generators: Create Custom Browser Add-ons with Zero Coding Experience
Discover the 2026 best AI Chrome extension generators on XIX.AI. Our curated list features top-rated, must-try tools that let you create custom browser add-ons with zero coding. Compare free vs paid options, see real-world tests, and unlock your productivity. Explore the latest rankings and find your perfect tool today!
10 tools
xix.ai
Text-to-speech
Best AI Multilingual TTS: Generate Authentic Native-Accent Speech in 50+ Languages
Discover the 2026 best AI multilingual TTS tools for authentic native-accent speech in 50+ languages. Explore our top-rated, curated rankings with free vs paid comparisons and real-world tests. Find your perfect voice tool on XIX.AI and unlock global communication today.
10 tools
xix.ai
Meeting Assistant
Best AI Meeting Automation Tools for Smarter and Faster Collaboration
Discover the 2026 latest top-rated AI meeting automation tools for smarter, faster collaboration. Our curated list features powerful, game-changing solutions to automate notes, summaries, and action items. Compare free vs paid options with real-world tests and weekly updated rankings. Unlock peak team productivity. Explore the best picks now at XIX.AI.
10 tools
xix.ai
Prompt
AI Prompts for Infrastructure-as-Code: Deploy Terraform & Docker Configurations Safely
Discover the 2026 latest top-rated AI prompts for Infrastructure-as-Code. XIX.AI's curated selection helps you safely deploy Terraform & Docker configurations, automate cloud setups, and boost DevOps productivity. Compare free vs paid options with real-world tests. Explore now and unlock your AI edge.
10 tools
xix.ai
Comments (1)
0/500
![CharlesYoung]()
Les mises à jour de politique me rappellent que les IA comme Claude sont des outils qui peuvent avoir des utilisations à double tranchant. 🔧 C'est rassurant de voir les entreprises prendre des initiatives, mais une réglementation internationale ne serait-elle pas plus efficace contre certains dangers ? En tout cas, voilà un pas dans la bonne direction.
Anthropic has updated the usage policy for its Claude AI chatbot to address growing safety concerns. Alongside stricter cybersecurity rules, the company now explicitly outlines categories of dangerous weapons that must not be developed using Claude.
While the official announcement post does not explicitly highlight the changes to its weapons policy, a comparison between the old and new usage guidelines reveals a significant update. Previously, Anthropic broadly banned using Claude to "produce, modify, design, market, or distribute weapons, explosives, dangerous materials or other systems designed to cause harm to or loss of human life." The revised policy expands this by specifically prohibiting the development of high-yield explosives, as well as biological, nuclear, chemical, and radiological (CBRN) weapons.
In May, Anthropic introduced "AI Safety Level 3" protections concurrent with the launch of its new Claude Opus 4 model. These safeguards are intended to make the model more resistant to jailbreaking and to help prevent it from aiding in the creation of CBRN weapons.
In its announcement, Anthropic also recognizes the risks associated with agentic AI tools like Computer Use—which allows Claude to control a user's computer—and Claude Code, which integrates Claude directly into a developer's terminal. "These powerful capabilities introduce new risks, including potential for scaled abuse, malware creation, and cyber attacks," the company states.
Related
- Anthropic launches new Claude service for military and intelligence use
- Google scraps promise not to develop AI weapons
The AI startup is addressing these potential threats by adding a new section titled "Do Not Compromise Computer or Network Systems" to its usage policy. This section includes rules against using Claude to discover or exploit vulnerabilities, create or distribute malware, develop tools for denial-of-service attacks, and similar activities.
Additionally, Anthropic is easing its policy regarding political content. Instead of banning all content creation related to political campaigns and lobbying, the company will now only prohibit using Claude for "use cases that are deceptive or disruptive to democratic processes, or involve voter and campaign targeting." Anthropic also clarified that its requirements for all "high-risk" use cases—which apply when Claude is used to make recommendations to individuals or customers—are relevant only for consumer-facing scenarios, not for internal business use.
Barry Diller: Trust in Sam Altman irrelevant as AGI nears
Barry Diller, the billionaire media titan, does not believe OpenAI CEO Sam Altman is untrustworthy, despite recent reports suggesting otherwise. Speaking at the Wall Street Journal's "Future of Everything" conference this week, Diller defended Altman
YouTube expands AI deepfake detection to politicians, government officials, and journalists
On Tuesday, YouTube announced it is expanding its deepfake detection technology to a select group of government officials, political candidates, and journalists. The tool identifies AI-generated likenesses and lets pilot participants request the remo
The Real Difference: Not One Thing, but Another
Sometimes, things are not only one thing but also another. The phrase "It's not just this — it's that" has become so common in AI-generated writing that it now serves as more than a hint of synthetic content — it's nearly a certainty.That's why, when
Discover the 2026 best AI assistants for crafting epic xianxia & wuxia tales. XIX.AI's curated list features top-rated, game-changing tools to master cultivation progression and martial arts choreography. Compare free vs paid options with real-world tests. Unlock your creative potential and start writing today!
10 tools
xix.ai
Discover the 2026 best AI mobile app coding tools for Flutter & React Native. Our curated, top-rated list features powerful, game-changing solutions that generate cross-platform code from prompts. Compare free vs paid options with real-world tests. Unlock faster development and build better apps. Explore the rankings on XIX.AI now!
10 tools
xix.ai
Discover the 2026 best AI Chrome extension generators on XIX.AI. Our curated list features top-rated, must-try tools that let you create custom browser add-ons with zero coding. Compare free vs paid options, see real-world tests, and unlock your productivity. Explore the latest rankings and find your perfect tool today!
10 tools
xix.ai
Discover the 2026 best AI multilingual TTS tools for authentic native-accent speech in 50+ languages. Explore our top-rated, curated rankings with free vs paid comparisons and real-world tests. Find your perfect voice tool on XIX.AI and unlock global communication today.
10 tools
xix.ai
Discover the 2026 latest top-rated AI meeting automation tools for smarter, faster collaboration. Our curated list features powerful, game-changing solutions to automate notes, summaries, and action items. Compare free vs paid options with real-world tests and weekly updated rankings. Unlock peak team productivity. Explore the best picks now at XIX.AI.
10 tools
xix.ai
Discover the 2026 latest top-rated AI prompts for Infrastructure-as-Code. XIX.AI's curated selection helps you safely deploy Terraform & Docker configurations, automate cloud setups, and boost DevOps productivity. Compare free vs paid options with real-world tests. Explore now and unlock your AI edge.
10 tools
xix.ai
Les mises à jour de politique me rappellent que les IA comme Claude sont des outils qui peuvent avoir des utilisations à double tranchant. 🔧 C'est rassurant de voir les entreprises prendre des initiatives, mais une réglementation internationale ne serait-elle pas plus efficace contre certains dangers ? En tout cas, voilà un pas dans la bonne direction.
