
AI-generated summary
Anthropic is developing the Claude model as a company focused on AI security. The company regularly publishes risk reports and implements security measures to prevent misuse of its models.
After Jacob Coxon, who worked at the artificial intelligence company Anthropic, resigned from the company with the warning that artificial intelligence could turn into superhuman systems in the future, a remarkable report was published by the company.
In the report published while Coxon's statements were on the agenda, it was stated that cases were detected in which Anthropic's artificial intelligence models were used in a way that could help develop biological weapons.
BIOLOGICAL WEAPON RISK
In the report published by Anthropic on the misuse of artificial intelligence models, it was emphasized that one of the most serious risks is "biological misuse". The report noted concerns that advanced artificial intelligence models could achieve capabilities that could make existing pathogens more dangerous or help develop new pathogens.
The company reported that this year it blocked some activities of scientists who were conducting research that could help develop biological weapons using artificial intelligence models, including Claude. Anthropic did not disclose the names, countries, or biological agents of the researchers in question.
"IT IS AN EXTREMELY NICE SITUATION"
Jacob Klein, Anthropic's head of threat intelligence, said it's not always clear whether bioprobes are malicious or not.
"You don't see someone like in the comics saying, 'Hey, I want to build a bioweapon that will kill everyone.' It's extremely subtle," Klein said. he said.
The company pointed out that legitimate biological research can also contribute to the development of dangerous pathogens. It was stated that studies such as vaccine development can produce beneficial results, but the same techniques can also be used to make pathogens more dangerous.
RESEARCHERS' ACTIVITIES WERE BLOCKED
The report stated that a researcher working in an unsupported region used Claude to plan experiments on the adaptation of bird flu to mammals. In another case, it was stated that a state-supported researcher tried to redesign toxins in a computer environment.
Anthropic said it detected users trying to bypass access restrictions and "obscure the purpose of their research to evade our security measures." The company banned accounts linked to these activities.
In another case in May, a scientist reportedly asked Claude to help him apply for funding to conduct experiments on the chikungunya virus, which can cause severe pain and other symptoms lasting for months.
Anthropic found the case alarming as it found that part of the research was planned to be done at a military research institute.
"What we don't know is whether this research was conducted for use as a weapon. But it is concerning that a military institution is conducting gain-of-function research," Klein said. he said.
OTHER THREATS ARE ALSO INCLUDED IN ANTHROPIC'S REPORT
Anthropic's report for the last eight months also included cases where artificial intelligence was misused for different purposes besides biological weapons.
The report stated that actors suspected of having ties to the Chinese and Iranian governments are using artificial intelligence models to spy on dissidents and diaspora communities.
It has been alleged that Russian state media also used Claude to produce online propaganda under the guise of independent reporting. These activities included the spread of fabricated claims about the elections in Moldova.
Anthropic also disclosed that it had identified cases where Claude was used to work on software for the design and development of conventional weapons, including firearms, missiles, armed drones and bombs. The report detailed three cases from China, two from Russia and one from Yemen.
ATTENTION-DRAWING RESIGNATION FROM ANTHROPIC EMPLOYEE
Anthropic employee Jacob Coxon's statements about the future of artificial intelligence were also on the agenda.
Coxon announced his resignation from the company, arguing that artificial intelligence "could become superhuman systems that can gain real power and resources."
Coxon's post was viewed millions of times and had a wide impact.
AI outlook — possibilities, not facts
Anthropic may further tighten its AI security measures.
Likely · Within weeks
International debates about the risk of biological weapons may increase.
Possible · Within months

Apple has created a new reference point in the industry by providing IP68 water and dust resistance with its new foldable phone, iPhone Duo. Samsung Galaxy Z Fold 8 remains at the IP48 level and can withstand water at a depth of 1.5 meters for 30 minutes. iPhone Duo, on the other hand, offers protection up to a depth of 6 meters. This difference could reshape the competition in the durability standards of foldable phones.

Apple is preparing to introduce new MacBook Pro models, OLED screen iPad mini and updated HomePod devices in October. The company aims to increase performance with MacBook Pro and iMac models with M6 processors. Siri AI integration and touch-screen HomePod models are expected in the home product line. iPad mini will be updated with OLED screen and A19/A20 Pro chip.

The Personal Data Protection Authority (KVKK) emphasized the risk of processing biometric data of artificial intelligence applications used in the 80s nostalgia trend popular on social media. Users are advised to check privacy policies before uploading their photos and avoid undefined apps.

Anthropic published a report highlighting the risk of AI models reaching the ability to make existing pathogens more dangerous or produce new pathogens. The report shared two examples where the Claude model was used to plan experiments on bird flu and toxins, and the accounts of the relevant users were blocked. The company stated that it aims to start a dialogue about biological risks in the industry by sharing these samples without claiming malicious use.

Google has launched the Windows application that brings the Gemini artificial intelligence experience to the desktop. The application provides instant access with the Alt + Space shortcut and works integrated with the Google ecosystem. It offers support for Windows 10 and 11.

The f/1.48 maximum aperture on the main camera of the iPhone 18 Pro is considered an important innovation to improve the daily photography experience of average users without photography knowledge by improving low-light performance and providing a more natural bokeh effect.