#claudemythospreview — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #claudemythospreview, aggregated by home.social.
-
https://www.europesays.com/at/159803/ Apple untersucht von Claude Mythos entdecktes macOS-Problem #Anthropic #Apple #AT #Austria #Claude #ClaudeMythosPreview #IT #Mac #macOS #Mobiles #Österreich #Science #Science&Technology #Security #Technik #Technology #Wissenschaft #Wissenschaft&Technik
-
IMF Warns AI Has Made Cyber Risk a Financial Stability Threat
Forrester’s Allie Mellen on Preparing for a Mythos-Level Surge in Vulnerabilities Jennifer Lawinski • May 13, 2026 …
#Economy #AIcybersecurityfinancialsector #CISOAIsecurity #ClaudeMythosPreview #cyberstresstesting #financialstabilitycyberrisk #IMF #IMFcyberrisk #InternationalMonetaryFund #OpenAIDaybreak #patchmanagementAI #ProjectGlasswing
https://www.europesays.com/2987477/ -
AI Is Supercharging Cybercrime— And IMF Says Finance May Not Be Ready
Benzinga and Yahoo Finance LLC may earn commission or revenue on some items through the links below. The…
#Economy #ClaudeMythosPreview #cloudservices #ConnectInvest #Financialsystems #IMF #InternationalMonetaryFund #majoroperatingsystems #marketdisruption #ModeMobile #realestate #softwareplatforms #vulnerabilities
https://www.europesays.com/2984518/ -
https://www.europesays.com/ee/159498/ Andrei Hvostov: Hormuzi väin on muutunud Schrödingeri väinaks. Keegi ei saa enam aru, kas see on lahti või kinni #Antropic #BreakingNews #BreakingNews #ClaudeMythosPreview #EE #Eesti #EestiKeel #Estonia #Estonian #FeaturedNews #FeaturedNews #Headlines #HormuziVäin #Iraan #kass #LatestNews #LatestNews #Maailm #News #PopulaarseimadLood #Schrödinger #TopStories #TopStories #Trump #ÜldisedUudised #usa #Uudised #ViimasedUudised #World #WorldNews #WorldNews
-
Anthropic Unveils AI Model Capable of Exploiting Software Vulnerabilities
Anthropic has just unveiled an AI model that can expose and exploit software vulnerabilities, raising a crucial question: can a tool that reveals the weaknesses of our digital world be safely shared with that world? The company has taken a cautious approach, limiting access to this powerful model to…
#AiModel #SoftwareVulnerabilityExploitation #EmergingThreats #Anthropic #ClaudeMythosPreview
-
https://www.europesays.com/people/28453/ How Anthropic’s Pentagon fight became an unexpected growth engine #AICompany #Anthropic #ClaudeMythos #ClaudeMythosPreview #DarioAmodei #DefenseDepartment #PeteHegseth #ThePentagon
-
https://www.europesays.com/news/15688/ Wipe out a ‘civilization’? Minor stuff compared to what just happened in AI #AI #Anthropic #civilization #ClaudeAiSuperBrain #ClaudeMythosPreview #company #Headlines #Infrastructure #job #News #RomanYampolskiy #scenario #second #SingularProblem #StateGovernment #TopStories #warning #world
-
https://www.europesays.com/people/19753/ Wipe out a ‘civilization’? Minor stuff compared to what just happened in AI #AI #Anthropic #civilization #ClaudeAiSuperBrain #ClaudeMythosPreview #company #DarioAmodei #Infrastructure #job #RomanYampolskiy #scenario #second #SingularProblem #StateGovernment #warning #world
-
Anthropic's Claude Mythos Preview Bolsters Cybersecurity Leaders
Anthropic's Claude Mythos Preview is giving select cybersecurity giants a powerful edge, and we're exploring what this exclusive rollout means for the industry. By granting early access to this cutting-edge AI model, Project Glasswing is poised to reshape the cybersecurity landscape.
#AiDevelopment #ClaudeMythosPreview #Cybersecurity #EmergingThreats #GenerativeAi
-
Anthropic's Claude Mythos Preview Bolsters Cybersecurity Leaders
Anthropic's Claude Mythos Preview is giving select cybersecurity giants a powerful edge, and we're exploring what this exclusive rollout means for the industry. By granting early access to this cutting-edge AI model, Project Glasswing is poised to reshape the cybersecurity landscape.
#AiDevelopment #ClaudeMythosPreview #Cybersecurity #EmergingThreats #GenerativeAi
-
Anthropic's Claude Mythos Preview Bolsters Cybersecurity Leaders
Anthropic's Claude Mythos Preview is giving select cybersecurity giants a powerful edge, and we're exploring what this exclusive rollout means for the industry. By granting early access to this cutting-edge AI model, Project Glasswing is poised to reshape the cybersecurity landscape.
#AiDevelopment #ClaudeMythosPreview #Cybersecurity #EmergingThreats #GenerativeAi
-
qwant news | Sam Altman says OpenAI is resetting Codex usage limits “to celebrate 3M weekly codex users” and will reset them for every 1M new users until it reaches 10M
Anthropic unveiled “Project Glasswing,” a new cybersecurity effort that gives a select group of partners early, controlled access to its unreleased Claude Mythos Preview model. The frontier‑class AI is designed to locate software vulnerabilities more effectively than even the most skilled human researchers, reportedly finding thousands of high‑severity bugs—including zero‑day flaws in every major operating system and web browser. Anthropic chose to restrict the model’s release because it believes the technology is powerful enough to be weaponized if made publicly available.
The initiative assembles a multi‑vendor coalition that includes Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, Microsoft, NVIDIA, Palo Alto Networks, JPMorgan Chase, the Linux Foundation and others. Together, these companies will use Mythos Preview to scan critical codebases, assist open‑source maintainers, and develop safeguards that can block the model’s most dangerous outputs. Anthropic is also pledging up to $100 million in usage credits and additional funding for open‑source security projects to help offset the resource demands of the effort.
Anthropic’s leadership frames Project Glasswing as a “first step” toward a broader, industry‑wide response to the emerging AI‑driven cyber threat landscape. By giving trusted defenders a head start on patching vulnerabilities, the company hopes to prevent the rapid proliferation of AI‑generated exploits while buying time to devise longer‑term safety and governance mechanisms. Executives acknowledge the high stakes: if the model is mishandled, it could give a private entity unprecedented zero‑day knowledge across the software ecosystem, making coordinated governance and transparency essential to avoid a potential security “shakedown.”
Read more: https://www.techmeme.com/260408/p7
#samaltman #openai #anthropic #projectglasswing #claudemythospreview
-
qwant news | Sam Altman says OpenAI is resetting Codex usage limits “to celebrate 3M weekly codex users” and will reset them for every 1M new users until it reaches 10M
Anthropic unveiled “Project Glasswing,” a new cybersecurity effort that gives a select group of partners early, controlled access to its unreleased Claude Mythos Preview model. The frontier‑class AI is designed to locate software vulnerabilities more effectively than even the most skilled human researchers, reportedly finding thousands of high‑severity bugs—including zero‑day flaws in every major operating system and web browser. Anthropic chose to restrict the model’s release because it believes the technology is powerful enough to be weaponized if made publicly available.
The initiative assembles a multi‑vendor coalition that includes Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, Microsoft, NVIDIA, Palo Alto Networks, JPMorgan Chase, the Linux Foundation and others. Together, these companies will use Mythos Preview to scan critical codebases, assist open‑source maintainers, and develop safeguards that can block the model’s most dangerous outputs. Anthropic is also pledging up to $100 million in usage credits and additional funding for open‑source security projects to help offset the resource demands of the effort.
Anthropic’s leadership frames Project Glasswing as a “first step” toward a broader, industry‑wide response to the emerging AI‑driven cyber threat landscape. By giving trusted defenders a head start on patching vulnerabilities, the company hopes to prevent the rapid proliferation of AI‑generated exploits while buying time to devise longer‑term safety and governance mechanisms. Executives acknowledge the high stakes: if the model is mishandled, it could give a private entity unprecedented zero‑day knowledge across the software ecosystem, making coordinated governance and transparency essential to avoid a potential security “shakedown.”
Read more: https://www.techmeme.com/260408/p7
#samaltman #openai #anthropic #projectglasswing #claudemythospreview
-
qwant news | Sam Altman says OpenAI is resetting Codex usage limits “to celebrate 3M weekly codex users” and will reset them for every 1M new users until it reaches 10M
Anthropic unveiled “Project Glasswing,” a new cybersecurity effort that gives a select group of partners early, controlled access to its unreleased Claude Mythos Preview model. The frontier‑class AI is designed to locate software vulnerabilities more effectively than even the most skilled human researchers, reportedly finding thousands of high‑severity bugs—including zero‑day flaws in every major operating system and web browser. Anthropic chose to restrict the model’s release because it believes the technology is powerful enough to be weaponized if made publicly available.
The initiative assembles a multi‑vendor coalition that includes Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, Microsoft, NVIDIA, Palo Alto Networks, JPMorgan Chase, the Linux Foundation and others. Together, these companies will use Mythos Preview to scan critical codebases, assist open‑source maintainers, and develop safeguards that can block the model’s most dangerous outputs. Anthropic is also pledging up to $100 million in usage credits and additional funding for open‑source security projects to help offset the resource demands of the effort.
Anthropic’s leadership frames Project Glasswing as a “first step” toward a broader, industry‑wide response to the emerging AI‑driven cyber threat landscape. By giving trusted defenders a head start on patching vulnerabilities, the company hopes to prevent the rapid proliferation of AI‑generated exploits while buying time to devise longer‑term safety and governance mechanisms. Executives acknowledge the high stakes: if the model is mishandled, it could give a private entity unprecedented zero‑day knowledge across the software ecosystem, making coordinated governance and transparency essential to avoid a potential security “shakedown.”
Read more: https://www.techmeme.com/260408/p7
#samaltman #openai #anthropic #projectglasswing #claudemythospreview
-
qwant news | Sam Altman says OpenAI is resetting Codex usage limits “to celebrate 3M weekly codex users” and will reset them for every 1M new users until it reaches 10M
Anthropic unveiled “Project Glasswing,” a new cybersecurity effort that gives a select group of partners early, controlled access to its unreleased Claude Mythos Preview model. The frontier‑class AI is designed to locate software vulnerabilities more effectively than even the most skilled human researchers, reportedly finding thousands of high‑severity bugs—including zero‑day flaws in every major operating system and web browser. Anthropic chose to restrict the model’s release because it believes the technology is powerful enough to be weaponized if made publicly available.
The initiative assembles a multi‑vendor coalition that includes Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, Microsoft, NVIDIA, Palo Alto Networks, JPMorgan Chase, the Linux Foundation and others. Together, these companies will use Mythos Preview to scan critical codebases, assist open‑source maintainers, and develop safeguards that can block the model’s most dangerous outputs. Anthropic is also pledging up to $100 million in usage credits and additional funding for open‑source security projects to help offset the resource demands of the effort.
Anthropic’s leadership frames Project Glasswing as a “first step” toward a broader, industry‑wide response to the emerging AI‑driven cyber threat landscape. By giving trusted defenders a head start on patching vulnerabilities, the company hopes to prevent the rapid proliferation of AI‑generated exploits while buying time to devise longer‑term safety and governance mechanisms. Executives acknowledge the high stakes: if the model is mishandled, it could give a private entity unprecedented zero‑day knowledge across the software ecosystem, making coordinated governance and transparency essential to avoid a potential security “shakedown.”
Read more: https://www.techmeme.com/260408/p7
#samaltman #openai #anthropic #projectglasswing #claudemythospreview
-
qwant news | Sam Altman says OpenAI is resetting Codex usage limits “to celebrate 3M weekly codex users” and will reset them for every 1M new users until it reaches 10M
Anthropic unveiled “Project Glasswing,” a new cybersecurity effort that gives a select group of partners early, controlled access to its unreleased Claude Mythos Preview model. The frontier‑class AI is designed to locate software vulnerabilities more effectively than even the most skilled human researchers, reportedly finding thousands of high‑severity bugs—including zero‑day flaws in every major operating system and web browser. Anthropic chose to restrict the model’s release because it believes the technology is powerful enough to be weaponized if made publicly available.
The initiative assembles a multi‑vendor coalition that includes Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, Microsoft, NVIDIA, Palo Alto Networks, JPMorgan Chase, the Linux Foundation and others. Together, these companies will use Mythos Preview to scan critical codebases, assist open‑source maintainers, and develop safeguards that can block the model’s most dangerous outputs. Anthropic is also pledging up to $100 million in usage credits and additional funding for open‑source security projects to help offset the resource demands of the effort.
Anthropic’s leadership frames Project Glasswing as a “first step” toward a broader, industry‑wide response to the emerging AI‑driven cyber threat landscape. By giving trusted defenders a head start on patching vulnerabilities, the company hopes to prevent the rapid proliferation of AI‑generated exploits while buying time to devise longer‑term safety and governance mechanisms. Executives acknowledge the high stakes: if the model is mishandled, it could give a private entity unprecedented zero‑day knowledge across the software ecosystem, making coordinated governance and transparency essential to avoid a potential security “shakedown.”
Read more: https://www.techmeme.com/260408/p7
#samaltman #openai #anthropic #projectglasswing #claudemythospreview
-
https://www.europesays.com/news/15017/ Anthropic Says Its Latest AI Model Is Too Powerful to Be Released #27YearOldVulnerability #Anthropic #capability #ClaudeMythosPreview #company #Cybersecurity #exploit #february #Headlines #LatestAiModel #model #mythos #MythosClassModel #News #part #researcher #safeguard #TopStories
-
Anthropic Withholds AI Model Over Misuse Fears
Anthropic has taken a bold step by withholding its latest artificial intelligence model from public release, citing concerns that its immense power could be misused. The company's new model, Claude Mythos Preview, has pushed the boundaries of automated capability, but Anthropic is taking a cautious approach to protect against potential risks.
https://osintsights.com/anthropic-withholds-ai-model-over-misuse-fears
#Anthropic #ArtificialIntelligence #AiModel #ClaudeMythosPreview #Cybersecurity
-
Anthropic Withholds AI Model Over Misuse Fears
Anthropic has taken a bold step by withholding its latest artificial intelligence model from public release, citing concerns that its immense power could be misused. The company's new model, Claude Mythos Preview, has pushed the boundaries of automated capability, but Anthropic is taking a cautious approach to protect against potential risks.
https://osintsights.com/anthropic-withholds-ai-model-over-misuse-fears
#Anthropic #ArtificialIntelligence #AiModel #ClaudeMythosPreview #Cybersecurity
-
Anthropic Withholds AI Model Over Misuse Fears
Anthropic has taken a bold step by withholding its latest artificial intelligence model from public release, citing concerns that its immense power could be misused. The company's new model, Claude Mythos Preview, has pushed the boundaries of automated capability, but Anthropic is taking a cautious approach to protect against potential risks.
https://osintsights.com/anthropic-withholds-ai-model-over-misuse-fears
#Anthropic #ArtificialIntelligence #AiModel #ClaudeMythosPreview #Cybersecurity
-
Anthropic Withholds AI Model Over Misuse Fears
Anthropic has taken a bold step by withholding its latest artificial intelligence model from public release, citing concerns that its immense power could be misused. The company's new model, Claude Mythos Preview, has pushed the boundaries of automated capability, but Anthropic is taking a cautious approach to protect against potential risks.
https://osintsights.com/anthropic-withholds-ai-model-over-misuse-fears
#Anthropic #ArtificialIntelligence #AiModel #ClaudeMythosPreview #Cybersecurity
-
Anthropic Withholds AI Model Over Misuse Fears
Anthropic has taken a bold step by withholding its latest artificial intelligence model from public release, citing concerns that its immense power could be misused. The company's new model, Claude Mythos Preview, has pushed the boundaries of automated capability, but Anthropic is taking a cautious approach to protect against potential risks.
https://osintsights.com/anthropic-withholds-ai-model-over-misuse-fears
#Anthropic #ArtificialIntelligence #AiModel #ClaudeMythosPreview #Cybersecurity
-
undefined | Anthropic limits Mythos AI rollout over fears hackers could use model for cyberattacks
Anthropic announced the rollout of Claude Mythos Preview, an advanced AI model designed to spot software weaknesses and security flaws. The company is limiting access to the model as a safeguard against misuse, making it available initially to a select group of partners—including Apple, Google, Microsoft, Nvidia, Amazon Web Services, CrowdStrike, Palo Alto Networks, and more than 40 other firms—through a new cybersecurity program called Project Glasswing. According to Anthropic’s head of research product management, Dianne Penn, the initiative is intended to give cyber‑defenders a “head start” on increasingly important threats, and the firm has pledged up to $100 million in usage credits for participating companies.
Claude Mythos Preview leverages Anthropic’s strong coding and reasoning abilities to find bugs that were previously difficult to detect. In testing, the model uncovered a 27‑year‑old vulnerability in OpenBSD, an operating system renowned for its security focus, and identified other critical flaws across first‑party and open‑source systems. Although the model was not specifically trained for cybersecurity, its general‑purpose capabilities have proven effective, prompting Anthropic to keep it out of broad public release while it learns how to safely scale Mythos‑class models. The company is working with U.S. government officials to ensure responsible deployment and to avoid the risks associated with releasing such powerful technology into hostile hands.
Anthropic’s founders, who left OpenAI over safety concerns, see this effort as a concrete step toward a more secure internet. CEO Dario Amodei emphasized that getting the balance right could create a “fundamentally more secure” digital world, while warning that mishandling the technology could pose significant danger. By partnering with critical software providers and offering limited‑time, credit‑backed access, Anthropic hopes to demonstrate the defensive value of AI‑driven security tools and set precedents for responsible AI use in the cybersecurity arena.
Read more: undefined
-
undefined | Anthropic limits Mythos AI rollout over fears hackers could use model for cyberattacks
Anthropic announced the rollout of Claude Mythos Preview, an advanced AI model designed to spot software weaknesses and security flaws. The company is limiting access to the model as a safeguard against misuse, making it available initially to a select group of partners—including Apple, Google, Microsoft, Nvidia, Amazon Web Services, CrowdStrike, Palo Alto Networks, and more than 40 other firms—through a new cybersecurity program called Project Glasswing. According to Anthropic’s head of research product management, Dianne Penn, the initiative is intended to give cyber‑defenders a “head start” on increasingly important threats, and the firm has pledged up to $100 million in usage credits for participating companies.
Claude Mythos Preview leverages Anthropic’s strong coding and reasoning abilities to find bugs that were previously difficult to detect. In testing, the model uncovered a 27‑year‑old vulnerability in OpenBSD, an operating system renowned for its security focus, and identified other critical flaws across first‑party and open‑source systems. Although the model was not specifically trained for cybersecurity, its general‑purpose capabilities have proven effective, prompting Anthropic to keep it out of broad public release while it learns how to safely scale Mythos‑class models. The company is working with U.S. government officials to ensure responsible deployment and to avoid the risks associated with releasing such powerful technology into hostile hands.
Anthropic’s founders, who left OpenAI over safety concerns, see this effort as a concrete step toward a more secure internet. CEO Dario Amodei emphasized that getting the balance right could create a “fundamentally more secure” digital world, while warning that mishandling the technology could pose significant danger. By partnering with critical software providers and offering limited‑time, credit‑backed access, Anthropic hopes to demonstrate the defensive value of AI‑driven security tools and set precedents for responsible AI use in the cybersecurity arena.
Read more: undefined
-
undefined | Anthropic Teams Up With Its Rivals to Keep AI From Hacking Everything by Lily Hay Newman
Anthropic unveiled Claude Mythos Preview, a powerful new AI model designed for code and cybersecurity tasks, and announced Project Glasswing—a cross‑industry consortium aimed at addressing the security risks the model could pose. The initiative gathers more than 45 organizations, including Microsoft, Apple, Google, AWS, the Linux Foundation, Cisco, Nvidia, Broadcom and dozens of other tech, cybersecurity, critical‑infrastructure and financial firms. By granting these members private access to Mythos Preview before a public release, Anthropic hopes developers can test the model on their own systems, identify vulnerabilities, and develop mitigation strategies before the technology becomes widely available.
The consortium’s members emphasize that the effort is not about promoting Anthropic’s product but about preparing for a future where AI‑driven attack capabilities are commonplace within the next year or two. According to Anthropic’s frontier red‑team lead Logan Graham, Mythos Preview can already perform tasks traditionally done by senior security researchers—discovering vulnerabilities, generating exploit chains, conducting penetration testing, assessing endpoints, hunting misconfigurations, and analyzing binaries without source code. Such capabilities could accelerate both defensive and offensive cyber operations, making coordinated vulnerability disclosure and careful release procedures essential to prevent the model from becoming a “meaningfully accelerant for attackers.”
Representatives from participating companies expressed strong support for the collaborative approach. Google’s VP of security engineering Heather Adkins highlighted AI’s new challenges and opportunities in cyber defense, while Microsoft’s global CISO Igor Tsyganskiy noted that early access to Mythos Preview enables the early identification and mitigation of risks at scale. Anthropic’s Dario Amodei stressed that more powerful models will emerge from multiple sources, underscoring the need for a collective response. Graham concluded that Project Glasswing must grow beyond a handful of firms to become a broader, sustained effort that can answer the pressing questions surrounding AI‑driven cybersecurity.
Read more: undefined
#business/artificialintelligence #security/cyberattacksandhacks #anthropic #claudemythospreview #projectglasswing
-
undefined | Anthropic Teams Up With Its Rivals to Keep AI From Hacking Everything by Lily Hay Newman
Anthropic unveiled Claude Mythos Preview, a powerful new AI model designed for code and cybersecurity tasks, and announced Project Glasswing—a cross‑industry consortium aimed at addressing the security risks the model could pose. The initiative gathers more than 45 organizations, including Microsoft, Apple, Google, AWS, the Linux Foundation, Cisco, Nvidia, Broadcom and dozens of other tech, cybersecurity, critical‑infrastructure and financial firms. By granting these members private access to Mythos Preview before a public release, Anthropic hopes developers can test the model on their own systems, identify vulnerabilities, and develop mitigation strategies before the technology becomes widely available.
The consortium’s members emphasize that the effort is not about promoting Anthropic’s product but about preparing for a future where AI‑driven attack capabilities are commonplace within the next year or two. According to Anthropic’s frontier red‑team lead Logan Graham, Mythos Preview can already perform tasks traditionally done by senior security researchers—discovering vulnerabilities, generating exploit chains, conducting penetration testing, assessing endpoints, hunting misconfigurations, and analyzing binaries without source code. Such capabilities could accelerate both defensive and offensive cyber operations, making coordinated vulnerability disclosure and careful release procedures essential to prevent the model from becoming a “meaningfully accelerant for attackers.”
Representatives from participating companies expressed strong support for the collaborative approach. Google’s VP of security engineering Heather Adkins highlighted AI’s new challenges and opportunities in cyber defense, while Microsoft’s global CISO Igor Tsyganskiy noted that early access to Mythos Preview enables the early identification and mitigation of risks at scale. Anthropic’s Dario Amodei stressed that more powerful models will emerge from multiple sources, underscoring the need for a collective response. Graham concluded that Project Glasswing must grow beyond a handful of firms to become a broader, sustained effort that can answer the pressing questions surrounding AI‑driven cybersecurity.
Read more: undefined
#business/artificialintelligence #security/cyberattacksandhacks #anthropic #claudemythospreview #projectglasswing
-
undefined | Anthropic Teams Up With Its Rivals to Keep AI From Hacking Everything by Lily Hay Newman
Anthropic unveiled Claude Mythos Preview, a powerful new AI model designed for code and cybersecurity tasks, and announced Project Glasswing—a cross‑industry consortium aimed at addressing the security risks the model could pose. The initiative gathers more than 45 organizations, including Microsoft, Apple, Google, AWS, the Linux Foundation, Cisco, Nvidia, Broadcom and dozens of other tech, cybersecurity, critical‑infrastructure and financial firms. By granting these members private access to Mythos Preview before a public release, Anthropic hopes developers can test the model on their own systems, identify vulnerabilities, and develop mitigation strategies before the technology becomes widely available.
The consortium’s members emphasize that the effort is not about promoting Anthropic’s product but about preparing for a future where AI‑driven attack capabilities are commonplace within the next year or two. According to Anthropic’s frontier red‑team lead Logan Graham, Mythos Preview can already perform tasks traditionally done by senior security researchers—discovering vulnerabilities, generating exploit chains, conducting penetration testing, assessing endpoints, hunting misconfigurations, and analyzing binaries without source code. Such capabilities could accelerate both defensive and offensive cyber operations, making coordinated vulnerability disclosure and careful release procedures essential to prevent the model from becoming a “meaningfully accelerant for attackers.”
Representatives from participating companies expressed strong support for the collaborative approach. Google’s VP of security engineering Heather Adkins highlighted AI’s new challenges and opportunities in cyber defense, while Microsoft’s global CISO Igor Tsyganskiy noted that early access to Mythos Preview enables the early identification and mitigation of risks at scale. Anthropic’s Dario Amodei stressed that more powerful models will emerge from multiple sources, underscoring the need for a collective response. Graham concluded that Project Glasswing must grow beyond a handful of firms to become a broader, sustained effort that can answer the pressing questions surrounding AI‑driven cybersecurity.
Read more: undefined
#business/artificialintelligence #security/cyberattacksandhacks #anthropic #claudemythospreview #projectglasswing
-
undefined | Anthropic Teams Up With Its Rivals to Keep AI From Hacking Everything by Lily Hay Newman
Anthropic unveiled Claude Mythos Preview, a powerful new AI model designed for code and cybersecurity tasks, and announced Project Glasswing—a cross‑industry consortium aimed at addressing the security risks the model could pose. The initiative gathers more than 45 organizations, including Microsoft, Apple, Google, AWS, the Linux Foundation, Cisco, Nvidia, Broadcom and dozens of other tech, cybersecurity, critical‑infrastructure and financial firms. By granting these members private access to Mythos Preview before a public release, Anthropic hopes developers can test the model on their own systems, identify vulnerabilities, and develop mitigation strategies before the technology becomes widely available.
The consortium’s members emphasize that the effort is not about promoting Anthropic’s product but about preparing for a future where AI‑driven attack capabilities are commonplace within the next year or two. According to Anthropic’s frontier red‑team lead Logan Graham, Mythos Preview can already perform tasks traditionally done by senior security researchers—discovering vulnerabilities, generating exploit chains, conducting penetration testing, assessing endpoints, hunting misconfigurations, and analyzing binaries without source code. Such capabilities could accelerate both defensive and offensive cyber operations, making coordinated vulnerability disclosure and careful release procedures essential to prevent the model from becoming a “meaningfully accelerant for attackers.”
Representatives from participating companies expressed strong support for the collaborative approach. Google’s VP of security engineering Heather Adkins highlighted AI’s new challenges and opportunities in cyber defense, while Microsoft’s global CISO Igor Tsyganskiy noted that early access to Mythos Preview enables the early identification and mitigation of risks at scale. Anthropic’s Dario Amodei stressed that more powerful models will emerge from multiple sources, underscoring the need for a collective response. Graham concluded that Project Glasswing must grow beyond a handful of firms to become a broader, sustained effort that can answer the pressing questions surrounding AI‑driven cybersecurity.
Read more: undefined
#business/artificialintelligence #security/cyberattacksandhacks #anthropic #claudemythospreview #projectglasswing