Anthropic Blocks Malicious AI Use, Updates Claude Models

Key Takeaways

• Anthropic released its third report on AI misuse since March 2025, detailing five cases of attempted biological weapon research.
• The company blocked requests to use Claude for gain-of-function research on the chikungunya virus and avian influenza.
• Anthropic updated its latest models with stronger safeguards, noting older 2025 models had weaker protections.
• Hackers used Claude to automate cyberattacks, stealing over 300,000 national identity records from a North African authority.
• Russian group Midnight Blizzard used Claude to hijack Wi-Fi networks and steal WhatsApp accounts.
• NetApp reported record first-quarter revenue of $2.03 billion driven by AI and cloud demand.
• Meta is manufacturing its custom Iris AI chip to reduce reliance on Nvidia hardware.
• Meta plans to reach a total data center computing capacity of 14 gigawatts by 2027.
• Anthropic scientist David Coxon resigned, warning that AI could get out of control by next year.
• OpenAI is campaigning for mandatory national regulations to address AI safety risks.

Anthropic blocks AI requests to make viruses harmful

Anthropic announced it stopped bad actors from using its AI models for dangerous activities like cyberattacks and biological weapon research. The company released its third report on AI misuse since March 2025, highlighting five specific cases where users tried to create biological threats. One notable attempt involved asking Claude to help write a grant for research on the chikungunya virus that could make it more transmissible. Anthropic added stronger safety rules to its latest models to prevent such misuse while noting that older models had weaker protections. The report warns that as AI becomes more powerful, the risk of misuse increases unless developers and governments act to keep these tools safe.

Anthropic blocks AI misuse for biological weapons

Anthropic stated it blocked efforts by bad actors to use its AI for malicious tasks including cyberattacks and research that could support biological weapons. The company added stronger safeguards to its latest models to restrict biological research that could be used to make weapons. This report is the third since March 2025 and includes examples of malicious code and AI prompts found during investigations. The report was published two days after a researcher resigned over concerns about AI safety and lack of responsibility in development. Between December 2025 and August 2026, researchers found misuse by spyware vendors and state-sponsored groups spreading propaganda.

Anthropic disrupts dozens of AI misuse attempts

Anthropic said it disrupted dozens of instances of potentially malicious use of its AI models since December 2025, including several cases of biological misuse. The report details five case studies where users tried to use Anthropic's models to support biological weapons development. In one case, a researcher outside the United States used Claude to study highly pathogenic avian influenza and how it adapts to mammals. Anthropic banned the accounts involved but noted it could not prove the researchers intended to cause harm. This report follows previous threat assessments and comes after former researcher Jacob Coxon spoke about concerns over powerful AI systems.

Anthropic blocks AI misuse for biological weapons

Anthropic said it blocked efforts by bad actors to use its artificial intelligence models for malicious activity such as cyberattacks and biological weapon research. The company added stronger safeguards in its latest models to restrict biological research that could also be used to make weapons. The report includes snippets of malicious code and AI prompts found during investigations and urges governments to prevent similar abuse. This announcement came two days after a researcher resigned over concerns that Anthropic and competitors are not acting responsibly in AI development. The report covers misuse by actors ranging from spyware vendors to state-sponsored groups spreading propaganda.

Anthropic blocks AI misuse for biological weapons

Anthropic stated it has blocked efforts by bad actors to use its artificial intelligence models for malicious activity such as cyberattacks and biological weapon research. The company added stronger safeguards in its latest models to restrict biological research that could also be used to make weapons. The report includes snippets of malicious code and AI prompts found during investigations and urges governments to prevent similar abuse. Anthropic noted that as models become more capable, their risks will increase unless developers and society act to make them safer. The company is planning an initial public offering this fall.

Anthropic Blocks AI Requests for Dangerous Virus Research

Anthropic released a report stating it blocked misuse of its AI models that could have supported biological weapons. The company found requests asking for help with gain-of-function research on the chikungunya virus, which aims to make the pathogen more harmful. Anthropic updated its newer models, like Claude Fable 5, with stronger safeguards to stop such dual-use research. Older models from 2025 had weaker protections because they were less capable of assisting in complex dangerous tasks. The company also shared details about groups creating fake social media accounts to spread political views from various countries.

Hackers Use Claude AI to Automate Cyberattacks and Evade Detection

A new report from Anthropic shows hackers used Claude AI agents to automate cyberattacks and create new security flaws. Groups like GTG-20006 used the AI to build malware that could change itself to avoid being detected by security software. These attackers targeted government agencies, drone manufacturers, and stole over 300,000 national identity records from a North African authority. Other criminals used the AI to scan millions of apps for secrets and steal payment card data to sell online. The findings cover attacks that happened between December 2025 and August 2026.

Russian Hackers Used Claude AI to Automate Malware Evasion

Anthropic reported that Russian hackers used Claude AI to speed up their ability to hide malware from defenders. The group, known as Midnight Blizzard, targeted over 20 organizations including Ukrainian and European government bodies. They used AI to hijack hotel Wi-Fi networks and take over WhatsApp accounts to steal private messages without the users knowing. Some attackers also tried to steal API keys from AI vendors to launch more attacks against other companies. Anthropic disrupted these operations and used the information to improve its own safety measures.

NetApp Stock Rallies As AI Demand Fuels Big Earnings Beat

NetApp stock rose by 7.98 percent after the company reported strong earnings driven by demand for AI and cloud services. The firm recorded record first-quarter revenue of $2.03 billion and earnings per share of $2.58. Management predicted second-quarter revenue between $2.025 billion and $2.175 billion, which is higher than what experts expected. Analysts raised their price targets for the stock because of this broad growth and improved profit margins. NetApp expects to grow revenue significantly through fiscal 2027 as more companies adopt its hybrid-cloud technology.

HPE Stock Rallies As AI And Networking Guidance Soars

Hewlett Packard Enterprise stock jumped 9.6 percent after investors praised its strong growth in AI and networking. The company reported record third-quarter revenue of $12.2 billion, a 34 percent increase from the previous year. Management raised its earnings guidance for the full year and expects revenue to grow by 34 to 37 percent. Analysts from Bank of America and Truist increased their price targets and kept their buy ratings on the stock. The company also forecasted strong cash flow and continued expansion in its cloud and AI business segments.

Senate probes AI safety after industry warnings

US lawmakers are pushing for new rules to govern artificial intelligence following recent safety concerns. Anthropic scientist David Coxon recently left the company, warning that AI could get out of control by next year. Another expert named Hubinger stated there is a greater than 10 percent chance AI could kill all humans within the next decade. OpenAI is now campaigning for mandatory national regulations to address these growing risks.

Overture brings paper AI classes to Swampscott

Overture is launching a new program in Swampscott this fall that teaches AI literacy using paper and pencil instead of screens. Co-founder Aspen Buckingham wants students to become responsible users and critical thinkers without excessive screen time. The program allows teachers to control technology access while students write down their ideas and design characters on paper. This approach helps fill a gap in education where many children lack guidance on how to safely use AI tools.

Guide to top AI speakers for business events

A new guide by Scott Steinberg helps event planners find the best artificial intelligence speakers for their meetings. These experts discuss how AI will change business, leadership, and the workforce in the coming years. The guide answers common questions about automation, risks, and how organizations can use AI responsibly. It aims to help leaders move past fear and hype to take informed action on this major technological shift.

Meta starts making its own Iris AI chip

Meta has begun manufacturing its custom Iris AI chip to reduce reliance on Nvidia hardware. The company plans to reach a total data center computing capacity of 14 gigawatts by 2027. This new chip is designed specifically for Meta's recommendation engines and generative AI features used on Facebook and Instagram. The move is part of a larger plan to release four different chips over the next two years.

Cerebras memory chip threatens Micron stock

Cerebras has developed a new memory technology called STT-MRAM that could replace traditional memory chips used in AI. This development has caused Micron shares to drop after the company had seen a historic surge of over 150 percent. Some analysts believe Cerebras stock could reach $150 per share, while others argue the technology is still too early. The debate continues over whether the AI memory trade has strong future potential or is running out of steam.

Moonshot AI aims for 2 billion dollars yearly revenue

Moonshot AI, the company behind the popular Kimi model, plans to earn 2 billion dollars in annual revenue. The Chinese lab hopes to turn its open-weight model into a major sales driver. Although recent usage numbers for its K3 model have dropped slightly, the company remains confident in its growth potential. Current reports suggest that Moonshot's projected earnings are still much lower than those of OpenAI and Anthropic. Despite these financial goals, the company faces criticism over its model development practices which some say are controversial or illegal.

Akamai and MuleSoft join forces to secure AI agents

Akamai and MuleSoft announced a new partnership on September 10, 2026, to improve security for autonomous AI systems. This collaboration helps bridge the gap between security policies and AI orchestration by connecting their platforms. The deal addresses a major problem where 87% of organizations faced API security incidents in 2025. The system creates a feedback loop where MuleSoft shares data with Akamai, which then sends threat scores back to MuleSoft. More than 20 customers are already using this integrated solution to stop shadow AI and data theft. Future updates planned for late 2026 will include new onboarding features and expanded security tools for AI environments.

Sources

• AI for bioweapons? Anthropic says Claude was asked to support making virus more ‘harmful’
• Anthropic says it blocked misuse of its AI that could have supported biological weapons
• Anthropic says it blocked potential AI bioweapon misuse
• Anthropic says it blocked misuse of its AI that could have supported biological weapons
• Anthropic says it blocked misuse of its AI that could have supported biological weapons
• Anthropic says it blocked misuse of its AI that could have supported biological weapons
• Hackers Use Claude AI Agents to Automate Cyberattacks, Develop 0-Days and Evade Detection
• Anthropic Says Russian Hackers Used Claude AI to Automate Malware Evasion
• NetApp Stock Rallies As AI Demand Fuels Big Earnings Beat
• HPE Stock Rallies As AI And Networking Guidance Soars
• OpenAI’s Hugging Face Hack Triggers Senate Probe
• Learning AI Without the Screen: Overture Comes to Swampscott This Fall
• Artificial Intelligence Speakers Guide by Keynote Scott Steinberg
• Meta Iris AI Chip Enters Production, Eyes 14GW by 2027
• Prediction: This AI Stock Could Be the Next One to Shock Wall Street
• Kimi-maker Moonshot AI targets $2 billion in annual revenue
• Akamai and MuleSoft Unify Runtime Enforcement Across Agent Fabric – Bridging the Gap Between Security Policy and AI Orchestration