<p>The US government has approved limited access to Anthropic's advanced Mythos 5 artificial intelligence model after weeks of negotiations over national security concerns. The decision marks a major policy shift after Washington temporarily blocked access to Mythos 5 and Fable 5 earlier this month, citing fears that the models could be "jailbroken" and misused for cyberattacks. The latest approval allows Anthropic to restore Mythos 5 to a small group of trusted cybersecurity partners, while broader restrictions on Fable 5 remain unchanged.</p><h2><strong>Why the US changed its decision</strong></h2><p>According to a letter from US Commerce Secretary Howard Lutnick, Anthropic made significant progress in addressing risks linked to the advanced AI model. Anthropic said it is now working to restore access for approved cyber defenders and infrastructure providers as quickly as possible. The company added that discussions with the US government will continue as it seeks wider access for Mythos 5 and hopes to make Fable 5 generally available again. The company has not disclosed exactly what technical or security changes were made to satisfy government officials.</p><p>&nbsp;</p><h2><strong>Why Mythos 5 drew attention</strong></h2><p>Mythos 5 is Anthropic's most advanced cybersecurity-focused AI model. It was designed to help researchers identify software vulnerabilities and strengthen digital defences, but officials worried the same capabilities could be exploited by malicious actors. Before restrictions were imposed, around 200 organisations, including Apple, Google, Microsoft, Nvidia and Cisco, reportedly had access to the model through Anthropic's invite-only Project Glasswing programme.</p><p>&nbsp;</p><h2><strong>Bigger impact on the AI industry</strong></h2><p>The dispute became one of the biggest government interventions in the AI industry so far. It also influenced rival OpenAI, which recently introduced a limited preview of GPT-5.6 following discussions with US officials. Anthropic says it will continue working with regulators to create a long-term framework for releasing powerful AI systems safely. The outcome could shape how future frontier AI models are launched, balancing rapid innovation with national security concerns.</p>

The US has approved limited access to Anthropic's Mythos 5 AI model after the company addressed national security concerns. The decision eases restrictions imposed earlier this month, while Fable 5 remains restricted as discussions with regulators continue.


The US government has approved limited access to Anthropic's advanced Mythos 5 artificial intelligence model after weeks of negotiations over national security concerns. The decision marks a major policy shift after Washington temporarily blocked access to Mythos 5 and Fable 5 earlier this month, citing fears that the models could be "jailbroken" and misused for cyberattacks. The latest approval allows Anthropic to restore Mythos 5 to a small group of trusted cybersecurity partners, while broader restrictions on Fable 5 remain unchanged. Why the US changed its decision According to a letter from US Commerce Secretary Howard Lutnick, Anthropic made significant progress in addressing risks linked to the advanced AI model. Anthropic said it is now working to restore access for approved cyber defenders and infrastructure providers as quickly as possible. The company added that discussions with the US government will continue as it seeks wider access for Mythos 5 and hopes to make Fable 5 generally available again. The company has not disclosed exactly what technical or security changes were made to satisfy government officials. Why Mythos 5 drew attention Mythos 5 is Anthropic's most advanced cybersecurity-focused AI model. It was designed to help researchers identify software vulnerabilities and strengthen digital defences, but officials worried the same capabilities could be exploited by malicious actors. Before restrictions were imposed, around 200 organisations, including Apple, Google, Microsoft, Nvidia and Cisco, reportedly had access to the model through Anthropic's invite-only Project Glasswing programme. Bigger impact on the AI industry The dispute became one of the biggest government interventions in the AI industry so far. It also influenced rival OpenAI, which recently introduced a limited preview of GPT-5.6 following discussions with US officials. Anthropic says it will continue working with regulators to create a long-term framework for releasing powerful AI systems safely. The outcome could shape how future frontier AI models are launched, balancing rapid innovation with national security concerns.

<p>The US has approved limited access to Anthropic's Mythos 5 AI model after the company addressed national security concerns. The decision eases restrictions imposed earlier this month, while Fable 5 remains restricted as discussions with regulators continue.</p>

Anthropic's most powerful AI model gets partial US approval. Here's what changed

<p>Anthropic has announced that Mythos-class artificial intelligence models will be made available to all customers in the coming weeks, marking a major step for one of the most closely watched AI systems in the cybersecurity industry. The announcement came alongside the launch of Anthropic's upgraded Claude Opus 4.8 model. While the new model attracted attention, many industry observers focused on a separate update confirming that Anthropic is preparing a broader rollout of AI systems with capabilities similar to Mythos. According to the company, safety protections are progressing quickly, allowing it to move closer to public availability.</p><h2><strong>What is Anthropic's Mythos AI?</strong></h2><p>Mythos is a specialised AI model designed to identify cybersecurity vulnerabilities in software systems. Unlike traditional AI chatbots that focus on answering questions or generating content, Mythos has been built to analyse software code, identify weaknesses and help organisations strengthen their digital security.</p><p>The model has attracted attention because of its reported ability to find large numbers of software vulnerabilities in a short period of time. Anthropic says Mythos can identify security flaws that may otherwise take human teams weeks or months to discover. Because of these capabilities, the company initially restricted access to the model through a programme called Project Glasswing, which was created to test the technology before a wider release.</p><h2><strong>Thousands of vulnerabilities already discovered</strong></h2><p>According to Anthropic, Mythos Preview has already helped uncover thousands of high-severity software vulnerabilities during testing. The company said some of these vulnerabilities affected major operating systems, web browsers and critical software infrastructure used worldwide. Anthropic reported that many organisations participating in Project Glasswing discovered hundreds of significant vulnerabilities in their own systems after using the model. Some participants also reported that their software security testing became more than ten times faster compared with traditional methods. One of the most notable examples came from Cloudflare, which said the model helped identify around 2,000 software bugs across important systems. According to Anthropic, nearly 400 of those vulnerabilities were classified as high or critical severity. The company also said Cloudflare found the model produced fewer false alarms than many traditional human-led testing methods.</p><h2><strong>Why cybersecurity experts are watching closely</strong></h2><p>The growing capabilities of AI systems in cybersecurity have generated both excitement and concern among industry experts. Supporters believe tools such as Mythos could significantly improve software security by helping companies identify weaknesses before cybercriminals exploit them. Faster detection of vulnerabilities could reduce the risk of major data breaches, ransomware attacks and disruptions to critical infrastructure. At the same time, experts have warned that powerful cybersecurity AI systems must be released carefully. If similar technology falls into the wrong hands, it could potentially be used to identify vulnerabilities for offensive purposes rather than defence.</p><p>This is one reason Anthropic created Project Glasswing, which includes participation from major technology firms such as Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike and Google.</p><h2><strong>Anthropic says safeguards are improving</strong></h2><p>Anthropic said the broader release is possible because the company is making progress on safety protections and monitoring systems. The company stated that it expects Mythos-class models to become available to customers in the coming weeks, although it has not yet confirmed whether the public version will be the exact Mythos model currently being tested or a closely related version with similar capabilities. The announcement reflects a broader trend across the AI industry, where companies are increasingly developing specialised models for fields such as cybersecurity, software engineering and scientific research.</p><p>The wider rollout of Mythos-class AI could become one of the most significant cybersecurity developments of the year. If Anthropic's claims are validated through broader deployment, organisations may gain access to tools capable of finding software vulnerabilities at a speed that was previously difficult to achieve. For businesses, governments and cybersecurity teams, the coming weeks will provide a clearer picture of how advanced AI systems can be used to strengthen digital defences while balancing safety and responsible deployment.</p><p>&nbsp;</p>

Anthropic has announced that Mythos-class artificial intelligence models will be made available to all customers in the coming weeks, marking a major step for one of the most closely watched AI systems in the cybersecurity industry.


Anthropic has announced that Mythos-class artificial intelligence models will be made available to all customers in the coming weeks, marking a major step for one of the most closely watched AI systems in the cybersecurity industry. The announcement came alongside the launch of Anthropic's upgraded Claude Opus 4.8 model. While the new model attracted attention, many industry observers focused on a separate update confirming that Anthropic is preparing a broader rollout of AI systems with capabilities similar to Mythos. According to the company, safety protections are progressing quickly, allowing it to move closer to public availability. What is Anthropic's Mythos AI? Mythos is a specialised AI model designed to identify cybersecurity vulnerabilities in software systems. Unlike traditional AI chatbots that focus on answering questions or generating content, Mythos has been built to analyse software code, identify weaknesses and help organisations strengthen their digital security. The model has attracted attention because of its reported ability to find large numbers of software vulnerabilities in a short period of time. Anthropic says Mythos can identify security flaws that may otherwise take human teams weeks or months to discover. Because of these capabilities, the company initially restricted access to the model through a programme called Project Glasswing, which was created to test the technology before a wider release. Thousands of vulnerabilities already discovered According to Anthropic, Mythos Preview has already helped uncover thousands of high-severity software vulnerabilities during testing. The company said some of these vulnerabilities affected major operating systems, web browsers and critical software infrastructure used worldwide. Anthropic reported that many organisations participating in Project Glasswing discovered hundreds of significant vulnerabilities in their own systems after using the model. Some participants also reported that their software security testing became more than ten times faster compared with traditional methods. One of the most notable examples came from Cloudflare, which said the model helped identify around 2,000 software bugs across important systems. According to Anthropic, nearly 400 of those vulnerabilities were classified as high or critical severity. The company also said Cloudflare found the model produced fewer false alarms than many traditional human-led testing methods. Why cybersecurity experts are watching closely The growing capabilities of AI systems in cybersecurity have generated both excitement and concern among industry experts. Supporters believe tools such as Mythos could significantly improve software security by helping companies identify weaknesses before cybercriminals exploit them. Faster detection of vulnerabilities could reduce the risk of major data breaches, ransomware attacks and disruptions to critical infrastructure. At the same time, experts have warned that powerful cybersecurity AI systems must be released carefully. If similar technology falls into the wrong hands, it could potentially be used to identify vulnerabilities for offensive purposes rather than defence. This is one reason Anthropic created Project Glasswing, which includes participation from major technology firms such as Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike and Google. Anthropic says safeguards are improving Anthropic said the broader release is possible because the company is making progress on safety protections and monitoring systems. The company stated that it expects Mythos-class models to become available to customers in the coming weeks, although it has not yet confirmed whether the public version will be the exact Mythos model currently being tested or a closely related version with similar capabilities. The announcement reflects a broader trend across the AI industry, where companies are increasingly developing specialised models for fields such as cybersecurity, software engineering and scientific research. The wider rollout of Mythos-class AI could become one of the most significant cybersecurity developments of the year. If Anthropic's claims are validated through broader deployment, organisations may gain access to tools capable of finding software vulnerabilities at a speed that was previously difficult to achieve. For businesses, governments and cybersecurity teams, the coming weeks will provide a clearer picture of how advanced AI systems can be used to strengthen digital defences while balancing safety and responsible deployment.

<p>Anthropic has announced that Mythos-class artificial intelligence models will be made available to all customers in the coming weeks, marking a major step for one of the most closely watched AI systems in the cybersecurity industry.</p>

Anthropic’s most powerful Mythos-class AI model could soon reach all users

<p>Anthropic’s Mythos model has drawn global attention after reports suggested it could identify thousands of software vulnerabilities. The release led to concerns among governments, banks and technology firms about the possibility of a new phase of AI-driven cyberattacks.</p><p>However, cybersecurity experts say that the capabilities highlighted by Mythos are not entirely new. According to researchers, similar results can already be achieved using existing AI models, including those from Anthropic and OpenAI.&nbsp;</p><h2>What Mythos revealed about cybersecurity risks</h2><p>Mythos gained attention because of its ability to detect large numbers of vulnerabilities in software systems. Some of these issues had remained unnoticed for years, raising concerns about how widely such flaws may exist.</p><p>The model was released in a limited way to selected organisations, including major companies, to reduce the risk of misuse. This controlled rollout reflects the sensitivity of tools that can identify weaknesses in critical systems.&nbsp;</p><h2>Experts say similar capabilities already exist</h2><p>Despite the attention around Mythos, researchers say that the ability to detect vulnerabilities at scale has been possible for some time.</p><p>Cybersecurity firms have shown that by using a method known as orchestration, multiple AI models can work together to analyse code and find similar vulnerabilities. In tests, older models from both Anthropic and OpenAI were able to detect many of the same issues identified by Mythos.</p><p>Experts note that this suggests the main difference is not just the model itself, but how effectively tools are combined and used.&nbsp;</p><h2>What are ‘zero-day’ vulnerabilities</h2><p>One of the key concerns in cybersecurity is the detection of “zero-day” vulnerabilities. These are flaws in software that are unknown to developers and have not yet been fixed.</p><p>Such vulnerabilities can be exploited by attackers before patches are released, making them particularly risky. Researchers say current AI tools are already capable of identifying these weaknesses, which adds to concerns about how quickly they can be used.</p><p>&nbsp;</p><p>Growing concern among governments and companies</p><p>The release of Mythos has increased discussions around cybersecurity risks at the highest levels. Reports indicate that policymakers and industry leaders are considering stronger oversight of advanced AI systems.</p><p>At the same time, organisations are concerned about the balance between detecting vulnerabilities and fixing them. While AI can identify issues quickly, addressing them still requires time and human effort.</p><h2>Offence may have an early advantage</h2><p>Experts suggest that the current situation may favour attackers in the short term. While AI tools can help defenders identify risks, they also make it easier for malicious actors to discover and exploit vulnerabilities.</p><p>Cybersecurity professionals note that hackers already have the skills needed to use such techniques, and AI could reduce the time required to carry out attacks. This could lead to an increase in the number of attempts on systems that were previously not targeted.</p><h2>Challenges in fixing vulnerabilities</h2><p>Even before AI tools became widely available, organisations faced challenges in managing software vulnerabilities. Fixing issues often requires testing, updates and sometimes taking systems offline, which can be complex and time-consuming.</p><p>Experts say that while AI has increased the number of vulnerabilities being identified, tools to fix them automatically are still limited. This creates a gap between detection and resolution.</p><h2><strong>What this means for the future</strong></h2><p>The discussion around Mythos highlights a broader trend in cybersecurity. AI is making it easier to identify risks, but it is also raising new challenges in how those risks are managed.</p><p>Experts believe that collaboration between companies, governments and researchers will be important in addressing these issues. As AI tools continue to improve, the focus will likely shift towards building systems that can not only detect vulnerabilities but also help fix them efficiently.</p><p>For now, the key takeaway is that the cybersecurity risks linked to AI are not entirely new, but they are becoming more visible and more urgent to address.</p><p>&nbsp;</p>

Anthropic’s Mythos model has drawn global attention after reports suggested it could identify thousands of software vulnerabilities. The release led to concerns among governments, banks and technology firms about the possibility of a new phase of AI-driven cyberattacks.


Anthropic’s Mythos model has drawn global attention after reports suggested it could identify thousands of software vulnerabilities. The release led to concerns among governments, banks and technology firms about the possibility of a new phase of AI-driven cyberattacks. However, cybersecurity experts say that the capabilities highlighted by Mythos are not entirely new. According to researchers, similar results can already be achieved using existing AI models, including those from Anthropic and OpenAI. What Mythos revealed about cybersecurity risks Mythos gained attention because of its ability to detect large numbers of vulnerabilities in software systems. Some of these issues had remained unnoticed for years, raising concerns about how widely such flaws may exist. The model was released in a limited way to selected organisations, including major companies, to reduce the risk of misuse. This controlled rollout reflects the sensitivity of tools that can identify weaknesses in critical systems. Experts say similar capabilities already exist Despite the attention around Mythos, researchers say that the ability to detect vulnerabilities at scale has been possible for some time. Cybersecurity firms have shown that by using a method known as orchestration, multiple AI models can work together to analyse code and find similar vulnerabilities. In tests, older models from both Anthropic and OpenAI were able to detect many of the same issues identified by Mythos. Experts note that this suggests the main difference is not just the model itself, but how effectively tools are combined and used. What are ‘zero-day’ vulnerabilities One of the key concerns in cybersecurity is the detection of “zero-day” vulnerabilities. These are flaws in software that are unknown to developers and have not yet been fixed. Such vulnerabilities can be exploited by attackers before patches are released, making them particularly risky. Researchers say current AI tools are already capable of identifying these weaknesses, which adds to concerns about how quickly they can be used. Growing concern among governments and companies The release of Mythos has increased discussions around cybersecurity risks at the highest levels. Reports indicate that policymakers and industry leaders are considering stronger oversight of advanced AI systems. At the same time, organisations are concerned about the balance between detecting vulnerabilities and fixing them. While AI can identify issues quickly, addressing them still requires time and human effort. Offence may have an early advantage Experts suggest that the current situation may favour attackers in the short term. While AI tools can help defenders identify risks, they also make it easier for malicious actors to discover and exploit vulnerabilities. Cybersecurity professionals note that hackers already have the skills needed to use such techniques, and AI could reduce the time required to carry out attacks. This could lead to an increase in the number of attempts on systems that were previously not targeted. Challenges in fixing vulnerabilities Even before AI tools became widely available, organisations faced challenges in managing software vulnerabilities. Fixing issues often requires testing, updates and sometimes taking systems offline, which can be complex and time-consuming. Experts say that while AI has increased the number of vulnerabilities being identified, tools to fix them automatically are still limited. This creates a gap between detection and resolution. What this means for the future The discussion around Mythos highlights a broader trend in cybersecurity. AI is making it easier to identify risks, but it is also raising new challenges in how those risks are managed. Experts believe that collaboration between companies, governments and researchers will be important in addressing these issues. As AI tools continue to improve, the focus will likely shift towards building systems that can not only detect vulnerabilities but also help fix them efficiently. For now, the key takeaway is that the cybersecurity risks linked to AI are not entirely new, but they are becoming more visible and more urgent to address.

<p>Anthropic’s Mythos model has drawn global attention after reports suggested it could identify thousands of software vulnerabilities. The release led to concerns among governments, banks and technology firms about the possibility of a new phase of AI-driven cyberattacks.</p>

‘AI cyber risk’: Why existing AI models are already powerful

<p>Anthropic’s new Mythos model is changing how companies approach cybersecurity, with early results showing a sharp improvement in detecting software vulnerabilities. When the model was introduced in April 2026, it reportedly identified thousands of high-severity bugs, many of which needed to be fixed before the system could be made widely available. Now, researchers working on Mozilla Firefox have shared how this technology is already being used in practice.</p><h2><strong>AI finds long-hidden bugs in Firefox</strong></h2><p>According to Mozilla’s security team, Mythos has uncovered several serious vulnerabilities, including issues that had remained undetected for more than a decade. This marks a significant shift compared to earlier AI tools, which often produced large numbers of low-quality alerts and false positives.</p><p>&nbsp;</p><p>Researchers say that recent improvements in AI systems, especially those that can review and refine their own outputs, have made these tools more reliable. As a result, security teams are now able to focus on meaningful findings instead of filtering out incorrect reports.</p><h2><strong>Bug fixes rise sharply with AI support</strong></h2><p>The impact of AI tools is clearly visible in Firefox’s recent updates. In April 2026, the browser released 423 bug fixes, compared to just 31 fixes in the same month last year. This increase shows how AI is helping teams identify more issues in less time. Some of these bugs were highly complex, including vulnerabilities in Firefox’s sandbox system. These types of issues are difficult to detect because they require multiple steps, including creating test code and simulating attacks on secure parts of the browser.</p><p>&nbsp;</p><p>&nbsp;</p>

Bug fixes rise sharply with AI support

<h2><strong>AI helps find bugs, but humans still fix them</strong></h2><p>Despite these advances, human engineers remain central to the process. While AI can suggest possible fixes, the final code is still written and reviewed by developers.</p><p>Mozilla’s team has said that AI-generated patches often need refinement and cannot be directly applied. For now, the role of AI is mainly to assist in discovery rather than fully automate the repair process.</p><p>&nbsp;</p><h2><strong>A shift in cybersecurity approach</strong></h2><p>The use of tools like Mythos reflects a broader change in cybersecurity. Instead of relying only on manual testing or traditional scanning tools, companies are now using AI to analyse code more deeply and identify hidden risks.</p><p>However, experts also note that these tools could be used by both defenders and attackers. While AI can help fix vulnerabilities faster, it may also enable malicious actors to discover weaknesses more quickly. Dario Amodei has suggested that, if managed properly, these tools could improve overall security by helping organisations fix vulnerabilities before they are exploited.</p><p>&nbsp;</p><p>At the same time, Mozilla engineers have taken a cautious view, noting that the long-term impact is still unclear. They believe that while AI gives an advantage to defenders, the balance between attackers and defenders may continue to evolve.</p><h2><strong>What this means going forward</strong></h2><p>The use of AI in cybersecurity is growing rapidly, and tools like Mythos show how quickly capabilities are improving. The ability to detect complex and long-standing bugs could help make software more secure.</p><p>At the same time, the technology raises new questions about how security risks will be managed in the future. As AI tools become more widely available, companies will need to adapt their strategies to keep systems safe while making the most of these new capabilities</p><p>&nbsp;</p>

Anthropic’s new Mythos model is changing how companies approach cybersecurity, with early results showing a sharp improvement in detecting software vulnerabilities. 

Anthropic’s new Mythos model is changing how companies approach cybersecurity, with early results showing a sharp improvement in detecting software vulnerabilities. When the model was introduced in April 2026, it reportedly identified thousands of high-severity bugs, many of which needed to be fixed before the system could be made widely available. Now, researchers working on Mozilla Firefox have shared how this technology is already being used in practice. AI finds long-hidden bugs in Firefox According to Mozilla’s security team, Mythos has uncovered several serious vulnerabilities, including issues that had remained undetected for more than a decade. This marks a significant shift compared to earlier AI tools, which often produced large numbers of low-quality alerts and false positives. Researchers say that recent improvements in AI systems, especially those that can review and refine their own outputs, have made these tools more reliable. As a result, security teams are now able to focus on meaningful findings instead of filtering out incorrect reports. Bug fixes rise sharply with AI support The impact of AI tools is clearly visible in Firefox’s recent updates. In April 2026, the browser released 423 bug fixes, compared to just 31 fixes in the same month last year. This increase shows how AI is helping teams identify more issues in less time. Some of these bugs were highly complex, including vulnerabilities in Firefox’s sandbox system. These types of issues are difficult to detect because they require multiple steps, including creating test code and simulating attacks on secure parts of the browser.

<p>Anthropic’s new Mythos model is changing how companies approach cybersecurity, with early results showing a sharp improvement in detecting software vulnerabilities.&nbsp;</p>

‘Anthropic’s Mythos AI bug detection’: What it means for the future of software security

anthropics

Anthropic's, News on Anthropic's, stories on Anthropic's, articles on Anthropic's, Anthropic's stories at Wion, Anthropic's news, Anthropic's latest news

anthropics

Anthropic's most powerful AI model gets partial US approval. Here's what changed

Anthropic’s most powerful Mythos-class AI model could soon reach all users

‘AI cyber risk’: Why existing AI models are already powerful

How Anthropic’s Mythos has rewritten Firefox’s approach to cybersecurity