In one case, stolen credentials and zero-day vulnerabilities were combined into an attack path.<\/p>\n\n\n\n
The models were not acting with malicious intent. They were pursuing their assigned objective and independently discovered a route that their developers had not anticipated.<\/p>\n\n\n\n
Goal + capability + access + insufficient control = potentially unpredictable actions.<\/strong><\/p>\n\n\n\n Anthropic has reported comparable cases during its own cybersecurity evaluations. Testing environments that should have remained isolated were accidentally connected to the internet, and models compromised real organizations in three instances using methods including weak passwords and unsecured endpoints. <\/p>\n\n\n\n The circumstances differed, but the lesson was similar: when capable autonomous systems gain real-world access, strong boundaries and oversight become essential.<\/p>\n\n\n\n This raises questions relevant far beyond AI laboratories:<\/p>\n\n\n\n AI agents are becoming increasingly capable of performing tasks that previously required substantial human expertise, including reconnaissance, vulnerability discovery, exploitation and adaptation when an initial method fails.<\/p>\n\n\n\n A malicious agent could potentially scan infrastructure continuously, identify weaknesses, test attack paths and immediately adapt to unsuccessful attempts.<\/p>\n\n\n\n That creates a difficult speed gap for defenders. Security analysts cannot manually investigate every event as quickly as automated systems can act. But giving defensive AI unrestricted autonomy is not the answer either\u2014the same incidents demonstrate the risks of insufficient control.<\/p>\n\n\n\n This is where we see the role of PULZARIS Analyst AI<\/strong><\/a>, the first component of the broader PULZARIS AI security operations ecosystem.<\/p>\n\n\n\n PULZARIS Analyst AI supports SOC analysts throughout investigations by triaging and analyzing alerts, correlating entities and context, enriching findings with threat intelligence, evaluating risk, and helping produce analyst-ready reports and recommendations. <\/p>\n\n\n\n Its purpose is not to replace security professionals or allow AI to make every critical decision. Instead, repetitive and data-intensive analytical work can be accelerated while analysts retain responsibility for important actions.<\/p>\n\n\n\n The approach combines AI speed with: <\/p>\n\n\n\n The OpenAI and Anthropic incidents do not mean that AI has become an independent cybercriminal. They demonstrate something more immediate: advanced AI agents can increasingly discover routes and attack paths that humans did not explicitly provide.<\/p>\n\n\n\n As attackers move toward machine-speed operations, defenders will need to do the same\u2014but with stronger safeguards.<\/p>\n\n\n\n For SOC teams, solutions such as PULZARIS Analyst AI therefore represent more than another layer of automation. They are part of the shift toward AI-enabled security operations that can keep pace with emerging threats while keeping humans firmly in control. <\/p>\n\n\n\n Read the full article on our International subsidiary\u2019s website by clicking on the logo:<\/a><\/a><\/a><\/a><\/a><\/a><\/a><\/a><\/a><\/p>\n\n\n\n <\/p> ![\"\"](\"https:\/\/euroone.hu\/wp-content\/uploads\/2026\/08\/when_AI_walks_its_own_path_EO-1024x576.png\")
<\/figure>\n\n\n\n
Similar warnings from Anthropic<\/h2>\n\n\n\n\n
Cyberattacks are approaching machine speed<\/h2>\n\n\n\nPULZARIS Analyst AI: speed without giving up control<\/h2>\n\n\n\n
\n
Sensitive telemetry can also be obfuscated before LLM processing, while analysts maintain visibility over investigations and their conclusions.<\/p>\n\n\n\n![\"\"](\"https:\/\/euroone.hu\/wp-content\/uploads\/2026\/08\/AI_on_the_defensive_side_human_in_the_decision_loop_EO-1024x616.png\")
<\/figure>\n\n\n\n
The next cybersecurity race has already begun<\/h2>\n\n\n\n![\"\"](\"https:\/\/euroone.hu\/wp-content\/uploads\/2024\/10\/socwise-poweredArtboard-1@4x-100.jpg\")
<\/a><\/figure>\n\n\n\nDo you have a question? Are you interested in the solution? Contact our colleagues! <\/h3>\n\n\n\n
<\/ul><\/div>\n